a new one-sample test for goodness-of-fit

15
This article was downloaded by: [University of California, Riverside Libraries] On: 03 November 2014, At: 10:05 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Communications in Statistics - Theory and Methods Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/lsta20 A New One-Sample Test for Goodness-of-Fit Joe Damico a b a Office of Information Technology , The Ohio State University , Columbus, Ohio, USA b Ohio State University , Office of Information Technology , 1121 Kinnear Road, Columbus, OH 43212, USA Published online: 16 Aug 2006. To cite this article: Joe Damico (2005) A New One-Sample Test for Goodness-of-Fit, Communications in Statistics - Theory and Methods, 33:1, 181-193, DOI: 10.1081/STA-120026585 To link to this article: http://dx.doi.org/10.1081/STA-120026585 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions

Upload: joe

Post on 07-Mar-2017

214 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: A New One-Sample Test for Goodness-of-Fit

This article was downloaded by [University of California Riverside Libraries]On 03 November 2014 At 1005Publisher Taylor amp FrancisInforma Ltd Registered in England and Wales Registered Number 1072954 Registered office MortimerHouse 37-41 Mortimer Street London W1T 3JH UK

Communications in Statistics - Theory and MethodsPublication details including instructions for authors and subscription informationhttpwwwtandfonlinecomloilsta20

A New One-Sample Test for Goodness-of-FitJoe Damico a ba Office of Information Technology The Ohio State University Columbus Ohio USAb Ohio State University Office of Information Technology 1121 Kinnear RoadColumbus OH 43212 USAPublished online 16 Aug 2006

To cite this article Joe Damico (2005) A New One-Sample Test for Goodness-of-Fit Communications in Statistics - Theoryand Methods 331 181-193 DOI 101081STA-120026585

To link to this article httpdxdoiorg101081STA-120026585

PLEASE SCROLL DOWN FOR ARTICLE

Taylor amp Francis makes every effort to ensure the accuracy of all the information (the ldquoContentrdquo) containedin the publications on our platform However Taylor amp Francis our agents and our licensors make norepresentations or warranties whatsoever as to the accuracy completeness or suitability for any purpose ofthe Content Any opinions and views expressed in this publication are the opinions and views of the authorsand are not the views of or endorsed by Taylor amp Francis The accuracy of the Content should not be reliedupon and should be independently verified with primary sources of information Taylor and Francis shallnot be liable for any losses actions claims proceedings demands costs expenses damages and otherliabilities whatsoever or howsoever caused arising directly or indirectly in connection with in relation to orarising out of the use of the Content

This article may be used for research teaching and private study purposes Any substantial or systematicreproduction redistribution reselling loan sub-licensing systematic supply or distribution in anyform to anyone is expressly forbidden Terms amp Conditions of access and use can be found at httpwwwtandfonlinecompageterms-and-conditions

A New One-Sample Test for Goodness-of-Fit

Joe Damico

Office of Information Technology The Ohio State UniversityColumbus Ohio USA

ABSTRACT

In this article we describe a new statistic It has a number of qualitiesthat recommend it for use as a one-sample test for goodness-of-fit

It is easy to describe and compute and so is useful as ateaching tool

It is a distribution-free statistic

Its distribution is skewed and it has a comparativelylarge range of values Therefore it can supply more criticalpoints that correspondto desired alpha levels

We can determine the 01 05 10 and 20 critical points forany large value of n by using a generalized formula

Correspondence Joe Damico Ohio State University Office of InformationTechnology 1121 Kinnear Road Columbus OH 43212 USA E-mail Damico1osuedu

181

DOI 101081STA-120026585 0361-0926 (Print) 1532-415X (Online)

Copyright 2004 by Marcel Dekker Inc wwwdekkercom

COMMUNICATIONS IN STATISTICS

Theory and Methods

Vol 33 No 1 pp 181ndash193 2004

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

We can extend the definition of this statistic to a two-samplesituation

It is a test that provides excellent power My results show thatthe power is on a par with the Cramer-Von Mises one-sampletest for goodness-of-fit

This article contains five sections as follows

1 Defining the new statistic A

2 Description of the tests of power

3 Tables of the distribution for the A statistic

4 Table summarizing the results of the power tests

5 A brief bibliography

Key Words Distribution-free Goodness-of-fit Greatest integerfunction Non-parametric test One-sample test Two-sample test

1 DEFINITION OF THE A STATISTIC

Let Xi ifrac14 1 n denote n observations from a population PxThat is the Xi are independent and identically distributed We wish totest H0 Px is a population with some specified probability density func-tion (pdf )

Begin by partitioning the range of the pdf into n equal and non-over-lapping intervals (That is the integral of the pdf over each of these inter-valsfrac14 1=n) Under the null hypothesis we would expect one observationfrom our sample to occur in each of these n intervals (The expectedvalues of the order statistics are i=(nthorn 1) where i is an integer in therange from 1 to n inclusive) Our one-sample A statistic is a measureof how much the actual observations deviate from this expectation

If you think of this situation as analogous to a row of boxes each ofwhich contains from 0 to n balls with the total of all balls equal to thetotal number of boxes (n) then you may think of computing this statisticas equivalent to computing the minimum number of lsquolsquomovesrsquorsquo required toproduce a row of boxes with one and only one ball in each box And alsquolsquomoversquorsquo is defined as a move of one ball from one box to an adjacentbox An example may be useful

Suppose we have a random sample comprising the following sevenvalues 9088 9328 9351 1023 1044 1048 and 1053 We wish to testthe hypothesis that these seven values were drawn from a normal distri-bution with meanfrac14 100 and standard deviationfrac14 5 We compute the

182 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

z-scores and the corresponding values from the normal distribution(1824) 0034 (1344) 0089 (1298) 0097 (046) 0678 (088)0811 (096) 0832 and (106) 0855 We proceed as if we were testingthe null hypothesis that these seven decimal numbers (0034 through0855) came from a uniform distribution on the interval (000 100)We begin by defining seven intervals and determining the number ofobservations in each

(0000 0143) 3(0143 0286) 0(0286 0429) 0(0429 0571) 0(0571 0714) 1(0714 0857) 3(0857 1000) 0

We now determine the number of moves required to produce the lsquolsquogroundstatersquorsquo (ie one ball in each box)

1stmove

2ndmove

3rdmove

4thmove

5thmove

6thmove

(0000 0143) 3 2 2 1 1 1 1(0143 0286) 0 1 0 1 1 1 1(0286 0429) 0 0 1 1 1 1 1(0429 0571) 0 0 0 0 0 1 1(0571 0714) 1 1 1 1 2 1 1(0714 0857) 3 3 3 3 2 2 1(0857 1000) 0 0 0 0 0 0 1

So the computed value of the A statistic is 6 The probability under thenull hypothesis that the A statistic assumes a value gtfrac14 6 is 03868 Thisalpha level would generally not be considered significant and so the nullhypothesis would not be rejected

The mathematical description of the statistic is simply a summationof a number of terms We begin with a few definitions As already men-tioned the Xi ifrac14 1 n denote n observations from a population PxLet Si ifrac14 1 n denote the order statistics Let F(si) be the value of thecumulative distribution function for the ith order statistic Finally letGif( ) represent the greatest integer function We define A as follows

A frac14Xn

ifrac141

jGifethnFethsiTHORN thorn 1THORN ij

A New One-Sample Test for Goodness-of-Fit 183

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

So for our example

Gifeth7 0034thorn 1THORN frac14 1 j1 1j frac14 0

Gifeth7 0089thorn 1THORN frac14 1 j1 2j frac14 1

Gifeth7 0097thorn 1THORN frac14 1 j1 3j frac14 2

Gifeth7 0678thorn 1THORN frac14 5 j5 4j frac14 1

Gifeth7 0812thorn 1THORN frac14 6 j6 5j frac14 1

Gifeth7 0834thorn 1THORN frac14 6 j6 6j frac14 0

Gifeth7 0855thorn 1THORN frac14 6 j6 7j frac14 1

And the value of A is 0thorn 1thorn 2thorn 1thorn 1thorn 0thorn 1frac14 6 as before It should beobvious that the absolute values of the seven differences computed aboveactually correspond to the number of lsquolsquomovesrsquorsquo required to place each ofthe n order statistics into its corresponding lsquolsquoboxrsquorsquo Thus the third orderstatistic which lsquolsquooriginated in box 1rsquorsquo must be moved twice in order tolsquolsquoend up in box 3rsquorsquo

This statistic shares with the chi-square the concept of partitioningthe range of the probability distribution However it has the decidedadvantage of being useful in situations where we are able to obtain onlya small sample

In section three of this article I have published the exact distributionof the A statistic for values of nfrac14 2 3 4 5 6 and 7 For larger values of nI used computer simulations to generate the approximate distributions

After completing this work I was able to compute some formulaethat give excellent predictions of the critical values for large values of n(These formulae were computed by using stepwise regression techniques)

The resulting formulae (for alpha levelsfrac14 01 05 10 and 20) aregiven in the Table 1 The values computed from these formulae fornfrac14 20 30 40 50 60 70 80 90 and 100 are given in Table 2

Obviously these are good estimations of the values computedthrough the simulation and appearing in section three of this article

Another interesting fact is that we could easily extend the definitionof this statistic to test the two-sample problem in the following way

Let Xi ifrac14 1 m denote m observations from a population Pxand let F(x) be the cdf Let yj jfrac14 1 n denote n observations froma second population Py and let G(y) be the cdf We wish to test the nullhypothesis Ffrac14G against the broad alternative hypothesis F not frac14 GAssume that m is less than or equal to n Let tk kfrac14 1 (mthorn n)be the order statistics from the combined samples We construct thetwo-sample statistic A2 using the ranks (rirsquos) of the original xirsquos in the

184 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

combined sample

A2 frac14Xm

ifrac141

jGifethethriTHORN ethm=ethmthorn nthorn 1THORNTHORN thorn 1THORN ij

Example We have a sample of size 4 from the x population and asample of size 6 from the y population

The x observations (and ranks) are

089eth1THORN 097eth2THORN 125eth3THORN 498eth5THORN

The y observations (and ranks) are

134eth4THORN 678eth6THORN 714eth7THORN 834eth8THORN 912eth9THORN 934eth10THORN

We compute

Gifetheth1THORN eth4=11THORN thorn 1THORN frac14 1 j1 1j frac14 0

Gifetheth2THORN eth4=11THORN thorn 1THORN frac14 1 j1 2j frac14 1

Gifetheth3THORN eth4=11THORN thorn 1THORN frac14 2 j2 3j frac14 1

Gifetheth5THORN eth4=11THORN thorn 1THORN frac14 2 j2 4j frac14 2

Table 2

Values of n

Alpha levels 20 30 40 50 60 70 80 90 100

020 37 68 104 145 190 240 293 349 409010 45 82 127 177 232 292 357 426 499005 52 96 147 205 270 340 415 495 580001 67 123 190 265 348 438 535 639 748

Table 1

Alpha level Critical value

020 Gif(04081 n(3=2)thorn 1)010 Gif(04981 n(3=2)thorn 1)005 Gif(05797 n(3=2)thorn 1)001 Gif(07473 n(3=2)thorn 1)

A New One-Sample Test for Goodness-of-Fit 185

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

So the statistic A2 has the value 4 The probability under the nullhypothesis that the A2 statistic assumes a value gtfrac14 4 is 00286 So thenull hypothesis would be rejected The cumulative distribution functionfor this two-sample test statistic (with mfrac14 4 and nfrac14 6) looks like this

2 POWER OF THE A STATISTIC

In an article titled lsquolsquoEDF Statistics for Goodness-of-fit and SomeComparisonsrsquorsquo Stephens 1974 presented a chart comparing the powersof several different statistical tests The null hypothesis is that we havea uniform random number on the interval (0 1) There were sevenalternate distributions defined as follows

F FethxTHORN frac14 1 eth1 zTHORNk 0 ltfrac14 z ltfrac14 1 k frac14 15 2

G FethxTHORN frac14 2ethk1THORNzk 0 ltfrac14 z ltfrac14 5

FethxTHORN frac14 1 2ethk1THORNeth1 zTHORNk 5 ltfrac14 z ltfrac14 1 k frac14 15 2 3

H FethxTHORN frac14 05 2ethk1THORNeth5 zTHORNk 0 ltfrac14 z ltfrac14 5

FethxTHORN frac14 05thorn 2ethk1THORNethz 5THORNk 5 ltfrac14 z ltfrac14 1 k frac14 15 2

I computed the power of the A statistic in these same seven situationsThe results are shown in section four of this paper The power of theA statistic compares very favorably with both the Kolmogorov-SmirnovD statistic and the Cramer-von Mises W2 statistic

3 TABLES OF THE A STATISTIC

Tables of the exact distribution of A for sample sizes (n) rangingfrom 2 through 7 is shown in Table 3

The tables of the A statistic (Table 4) were computed using simula-tions The number of replications (N) is shown followed by the samplesize (n) and four critical values with the corresponding alpha levels

Value 0 1 2 3 4Cumulative density 1714 5143 8381 9714 1000

186 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

In each case the four alpha levels were selected to be near 020 010 005and 001

Here are some additional critical values for values of n from 8through 15 (Tables 5 and 6)

4 TABLES COMPARING POWER OF THE

A STATISTIC WITH OTHER

ONE-SAMPLE TESTS

As noted earlier the table of results comparing Kolmogorov-Smirnov D Cramer-von Mises W2 Kuiper V Watson U2 Anderson-Darling A2 Q (frac14Si ln zi) and Chi2 first appeared in an article byStephens (1974) in JASA I have simply added the column showing thepower of the A statistic as estimated by a number of simulations

Table 3

A nfrac14 2 nfrac14 3 nfrac14 4 nfrac14 5 nfrac14 6 nfrac14 7

0 0500 022222 009375 003840 001543 000611991 1000 066667 037500 019200 009259 004283932 092593 068750 044160 025720 013973773 100000 087500 067200 046296 029273524 096094 082560 064472 046256235 099219 091520 077761 061317996 100000 096352 086703 073131097 098656 092438 081845158 099616 095945 088053459 099936 097990 0923662710 100000 099096 0952885311 099640 0972077912 099880 0984233913 099970 0991597214 099996 0995813215 100000 0998074216 0999198617 0999708518 0999912619 0999980620 0999997621 10000000

A New One-Sample Test for Goodness-of-Fit 187

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

10000 8 9 02267 1000 18 32 019611 01183 38 010113 00543 44 004917 00089 57 0009

10000 9 11 02074 1000 19 33 020413 01176 40 010216 00462 48 004920 00106 63 0010

10000 10 13 02008 10000 20 37 0197316 00972 45 0099319 00406 53 0047724 00090 66 00092

10000 11 15 02018 1000 21 40 019719 00860 49 009521 00522 56 005127 00111 75 0009

10000 12 17 02015 1000 22 43 020020 01162 51 010024 00507 59 004930 00099 77 0010

10000 13 19 02056 1000 23 47 019823 01038 56 010027 00484 66 004935 00097 83 0010

10000 14 22 01909 1200 24 48 0200827 00945 58 0098331 00497 67 0049240 00085 87 00092

10000 15 23 02125 10000 25 51 0199129 00972 61 0103134 00468 72 0051543 00089 93 00107

1000 16 26 0192 1600 26 53 0203131 0092 65 0101336 0046 77 0048148 0010 97 00094

1000 17 29 0193 1600 27 57 0196934 0102 69 0098840 0045 80 0047551 0008 104 00100

188 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4 Continued

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

1600 28 59 01981 1600 38 96 02019

73 01019 118 0101386 00500 135 00500110 00100 181 00100

1600 29 66 01975 1600 39 99 0201380 01013 120 0100092 00481 140 00494117 00100 170 00094

10000 30 67 02029 10000 40 103 0201181 01025 126 0100095 00490 148 00496126 00095 191 00099

1600 31 71 01969 10000 50 145 0194587 00981 177 00984101 00500 205 00496128 00100 265 00099

1600 32 74 02038 10000 60 190 0199991 01000 232 00986107 00488 270 00498135 00100 348 00093

1600 33 77 01963 10000 70 240 0201990 01006 292 01027108 00500 340 00488141 00094 438 00095

1600 34 82 01975 10000 80 293 01988101 00988 357 01015117 00488 415 00530149 00100 535 00112

10000 35 84 02036 10000 90 349 01982102 01019 426 01015120 00493 495 00530152 00103 639 00112

1600 36 91 02019 10000 100 409 01930113 00981 499 00955132 00488 580 00467166 00100 748 00089

1600 37 90 01988110 01000128 00488172 00100

A New One-Sample Test for Goodness-of-Fit 189

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

5

nfrac148

nfrac149

nfrac1410

nfrac1411

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

902267

11

02074

13

02008

15

02018

10

01657

12

01580

14

01606

16

01628

11

01183

13

01176

15

01260

17

01336

12

00810

14

00874

16

00972

18

01091

13

00543

15

00633

17

00729

19

00860

14

00347

16

00462

18

00542

20

00670

15

00227

17

00335

19

00406

21

00522

16

00151

18

00241

20

00298

22

00417

17

00089

19

00157

21

00221

23

00334

20

00106

22

00166

24

00257

21

00062

23

00126

25

00194

24

00090

26

00147

27

00111

28

00082

190 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

6

nfrac1412

nfrac1413

nfrac1414

nfrac1415

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

17

02015

19

02056

21

02168

23

02125

18

01677

20

01733

22

01909

24

01873

19

01407

21

01488

23

01657

25

01648

20

01162

22

01249

24

01431

26

01454

21

00929

23

01038

25

01267

27

01285

22

00755

24

00878

26

01100

28

01132

23

00615

25

00736

27

00945

29

00972

24

00507

26

00606

28

00806

30

00841

25

00416

27

00484

29

00691

31

00723

26

00315

28

00410

30

00581

32

00624

27

00240

29

00334

31

00497

33

00542

28

00175

30

00278

32

00435

34

00468

29

00134

31

00221

33

00360

35

00403

30

00099

32

00184

34

00293

36

00344

33

00145

35

00236

37

00296

34

00120

36

00200

38

00240

35

00097

37

00175

39

00206

38

00141

40

00163

39

00119

41

00141

40

00085

42

00118

43

00089

A New One-Sample Test for Goodness-of-Fit 191

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

(Table 7) The alternative distribution functions are described in sectiontwo of this paper The sample sizes (n) were set equal to 10 20 and 40The alpha level was set at 010

REFERENCES

Bradley J V (1968) Distribution-Free Statistical Tests EnglewoodCliffs NJ Prentice-Hall

Gibbons J D (1976) Nonparametric Methods for Quantitative AnalysisNew York Holt Rinehart and Winston

Gibbons J D (1985) Nonparametric Statistical Inference New YorkMarcel Dekker

Table 7

Alternative D W2 V U2 A2 Q Chi2 A

(nfrac14 10 alphafrac14 010)Fkfrac1415 023 027 018 019 024 043 ndash 027Fkfrac1420 054 060 035 035 058 ndash ndash 058Gkfrac1415 009 007 022 023 006 ndash ndash 007Gkfrac1420 009 007 040 044 006 ndash ndash 010Gkfrac1430 021 021 081 086 018 ndash ndash 029Hkfrac1415 ndash ndash ndash ndash ndash ndash ndash 014Hkfrac1420 ndash ndash ndash ndash ndash ndash ndash 021

(nfrac14 20 alphafrac14 010)Fkfrac1415 038 046 025 028 046 068 ndash 046Fkfrac1420 078 087 061 060 087 097 059 088Gkfrac1415 013 011 032 034 010 011 ndash 010Gkfrac1420 025 025 071 077 028 025 ndash 029Gkfrac1430 063 079 099 099 084 ndash ndash 083Hkfrac1415 025 020 036 037 028 ndash ndash 017Hkfrac1420 047 044 071 077 054 ndash ndash 036

(nfrac14 40 alphafrac14 010)Fkfrac1415 060 070 043 043 ndash 089 040 073Fkfrac1420 098 099 091 089 ndash 100 089 099Gkfrac1415 019 022 057 061 ndash ndash 039 023Gkfrac1420 056 072 096 098 ndash ndash 085 075Gkfrac1430 ndash ndash ndash ndash ndash ndash ndash 100Hkfrac1415 036 032 058 063 ndash ndash ndash 027Hkfrac1420 071 080 096 098 ndash ndash ndash 075

192 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Hollander M Wolfe D A (1973) Nonparametric Statistical Methods2nd ed New York Wiley

Stephens M A (1974) EDF statistics for goodness-of-fit and somecomparisons (In theory and methods) J Amer Stat Assoc Theoryand Methods Section September 69(347)730ndash737

A New One-Sample Test for Goodness-of-Fit 193

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Request PermissionOrder Reprints

Reprints of this article can also be ordered at

httpwwwdekkercomservletproductDOI101081STA120026585

Request Permission or Order Reprints Instantly

Interested in copying and sharing this article In most cases US Copyright Law requires that you get permission from the articlersquos rightsholder before using copyrighted content

All information and materials found in this article including but not limited to text trademarks patents logos graphics and images (the Materials) are the copyrighted works and other forms of intellectual property of Marcel Dekker Inc or its licensors All rights not expressly granted are reserved

Get permission to lawfully reproduce and distribute the Materials or order reprints quickly and painlessly Simply click on the Request Permission Order Reprints link below and follow the instructions Visit the US Copyright Office for information on Fair Use limitations of US copyright law Please refer to The Association of American Publishersrsquo (AAP) website for guidelines on Fair Use in the Classroom

The Materials are for your personal use only and cannot be reformatted reposted resold or distributed by electronic means or otherwise without permission from Marcel Dekker Inc Marcel Dekker Inc grants you the limited right to display the Materials only on your personal computer or personal wireless device and to copy and download single copies of such Materials provided that any copyright trademark or other notice appearing on such Materials is also retained by displayed copied or downloaded as part of the Materials and is not removed or obscured and provided you do not edit modify alter or enhance the Materials Please refer to our Website User Agreement for more details

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Page 2: A New One-Sample Test for Goodness-of-Fit

A New One-Sample Test for Goodness-of-Fit

Joe Damico

Office of Information Technology The Ohio State UniversityColumbus Ohio USA

ABSTRACT

In this article we describe a new statistic It has a number of qualitiesthat recommend it for use as a one-sample test for goodness-of-fit

It is easy to describe and compute and so is useful as ateaching tool

It is a distribution-free statistic

Its distribution is skewed and it has a comparativelylarge range of values Therefore it can supply more criticalpoints that correspondto desired alpha levels

We can determine the 01 05 10 and 20 critical points forany large value of n by using a generalized formula

Correspondence Joe Damico Ohio State University Office of InformationTechnology 1121 Kinnear Road Columbus OH 43212 USA E-mail Damico1osuedu

181

DOI 101081STA-120026585 0361-0926 (Print) 1532-415X (Online)

Copyright 2004 by Marcel Dekker Inc wwwdekkercom

COMMUNICATIONS IN STATISTICS

Theory and Methods

Vol 33 No 1 pp 181ndash193 2004

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

We can extend the definition of this statistic to a two-samplesituation

It is a test that provides excellent power My results show thatthe power is on a par with the Cramer-Von Mises one-sampletest for goodness-of-fit

This article contains five sections as follows

1 Defining the new statistic A

2 Description of the tests of power

3 Tables of the distribution for the A statistic

4 Table summarizing the results of the power tests

5 A brief bibliography

Key Words Distribution-free Goodness-of-fit Greatest integerfunction Non-parametric test One-sample test Two-sample test

1 DEFINITION OF THE A STATISTIC

Let Xi ifrac14 1 n denote n observations from a population PxThat is the Xi are independent and identically distributed We wish totest H0 Px is a population with some specified probability density func-tion (pdf )

Begin by partitioning the range of the pdf into n equal and non-over-lapping intervals (That is the integral of the pdf over each of these inter-valsfrac14 1=n) Under the null hypothesis we would expect one observationfrom our sample to occur in each of these n intervals (The expectedvalues of the order statistics are i=(nthorn 1) where i is an integer in therange from 1 to n inclusive) Our one-sample A statistic is a measureof how much the actual observations deviate from this expectation

If you think of this situation as analogous to a row of boxes each ofwhich contains from 0 to n balls with the total of all balls equal to thetotal number of boxes (n) then you may think of computing this statisticas equivalent to computing the minimum number of lsquolsquomovesrsquorsquo required toproduce a row of boxes with one and only one ball in each box And alsquolsquomoversquorsquo is defined as a move of one ball from one box to an adjacentbox An example may be useful

Suppose we have a random sample comprising the following sevenvalues 9088 9328 9351 1023 1044 1048 and 1053 We wish to testthe hypothesis that these seven values were drawn from a normal distri-bution with meanfrac14 100 and standard deviationfrac14 5 We compute the

182 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

z-scores and the corresponding values from the normal distribution(1824) 0034 (1344) 0089 (1298) 0097 (046) 0678 (088)0811 (096) 0832 and (106) 0855 We proceed as if we were testingthe null hypothesis that these seven decimal numbers (0034 through0855) came from a uniform distribution on the interval (000 100)We begin by defining seven intervals and determining the number ofobservations in each

(0000 0143) 3(0143 0286) 0(0286 0429) 0(0429 0571) 0(0571 0714) 1(0714 0857) 3(0857 1000) 0

We now determine the number of moves required to produce the lsquolsquogroundstatersquorsquo (ie one ball in each box)

1stmove

2ndmove

3rdmove

4thmove

5thmove

6thmove

(0000 0143) 3 2 2 1 1 1 1(0143 0286) 0 1 0 1 1 1 1(0286 0429) 0 0 1 1 1 1 1(0429 0571) 0 0 0 0 0 1 1(0571 0714) 1 1 1 1 2 1 1(0714 0857) 3 3 3 3 2 2 1(0857 1000) 0 0 0 0 0 0 1

So the computed value of the A statistic is 6 The probability under thenull hypothesis that the A statistic assumes a value gtfrac14 6 is 03868 Thisalpha level would generally not be considered significant and so the nullhypothesis would not be rejected

The mathematical description of the statistic is simply a summationof a number of terms We begin with a few definitions As already men-tioned the Xi ifrac14 1 n denote n observations from a population PxLet Si ifrac14 1 n denote the order statistics Let F(si) be the value of thecumulative distribution function for the ith order statistic Finally letGif( ) represent the greatest integer function We define A as follows

A frac14Xn

ifrac141

jGifethnFethsiTHORN thorn 1THORN ij

A New One-Sample Test for Goodness-of-Fit 183

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

So for our example

Gifeth7 0034thorn 1THORN frac14 1 j1 1j frac14 0

Gifeth7 0089thorn 1THORN frac14 1 j1 2j frac14 1

Gifeth7 0097thorn 1THORN frac14 1 j1 3j frac14 2

Gifeth7 0678thorn 1THORN frac14 5 j5 4j frac14 1

Gifeth7 0812thorn 1THORN frac14 6 j6 5j frac14 1

Gifeth7 0834thorn 1THORN frac14 6 j6 6j frac14 0

Gifeth7 0855thorn 1THORN frac14 6 j6 7j frac14 1

And the value of A is 0thorn 1thorn 2thorn 1thorn 1thorn 0thorn 1frac14 6 as before It should beobvious that the absolute values of the seven differences computed aboveactually correspond to the number of lsquolsquomovesrsquorsquo required to place each ofthe n order statistics into its corresponding lsquolsquoboxrsquorsquo Thus the third orderstatistic which lsquolsquooriginated in box 1rsquorsquo must be moved twice in order tolsquolsquoend up in box 3rsquorsquo

This statistic shares with the chi-square the concept of partitioningthe range of the probability distribution However it has the decidedadvantage of being useful in situations where we are able to obtain onlya small sample

In section three of this article I have published the exact distributionof the A statistic for values of nfrac14 2 3 4 5 6 and 7 For larger values of nI used computer simulations to generate the approximate distributions

After completing this work I was able to compute some formulaethat give excellent predictions of the critical values for large values of n(These formulae were computed by using stepwise regression techniques)

The resulting formulae (for alpha levelsfrac14 01 05 10 and 20) aregiven in the Table 1 The values computed from these formulae fornfrac14 20 30 40 50 60 70 80 90 and 100 are given in Table 2

Obviously these are good estimations of the values computedthrough the simulation and appearing in section three of this article

Another interesting fact is that we could easily extend the definitionof this statistic to test the two-sample problem in the following way

Let Xi ifrac14 1 m denote m observations from a population Pxand let F(x) be the cdf Let yj jfrac14 1 n denote n observations froma second population Py and let G(y) be the cdf We wish to test the nullhypothesis Ffrac14G against the broad alternative hypothesis F not frac14 GAssume that m is less than or equal to n Let tk kfrac14 1 (mthorn n)be the order statistics from the combined samples We construct thetwo-sample statistic A2 using the ranks (rirsquos) of the original xirsquos in the

184 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

combined sample

A2 frac14Xm

ifrac141

jGifethethriTHORN ethm=ethmthorn nthorn 1THORNTHORN thorn 1THORN ij

Example We have a sample of size 4 from the x population and asample of size 6 from the y population

The x observations (and ranks) are

089eth1THORN 097eth2THORN 125eth3THORN 498eth5THORN

The y observations (and ranks) are

134eth4THORN 678eth6THORN 714eth7THORN 834eth8THORN 912eth9THORN 934eth10THORN

We compute

Gifetheth1THORN eth4=11THORN thorn 1THORN frac14 1 j1 1j frac14 0

Gifetheth2THORN eth4=11THORN thorn 1THORN frac14 1 j1 2j frac14 1

Gifetheth3THORN eth4=11THORN thorn 1THORN frac14 2 j2 3j frac14 1

Gifetheth5THORN eth4=11THORN thorn 1THORN frac14 2 j2 4j frac14 2

Table 2

Values of n

Alpha levels 20 30 40 50 60 70 80 90 100

020 37 68 104 145 190 240 293 349 409010 45 82 127 177 232 292 357 426 499005 52 96 147 205 270 340 415 495 580001 67 123 190 265 348 438 535 639 748

Table 1

Alpha level Critical value

020 Gif(04081 n(3=2)thorn 1)010 Gif(04981 n(3=2)thorn 1)005 Gif(05797 n(3=2)thorn 1)001 Gif(07473 n(3=2)thorn 1)

A New One-Sample Test for Goodness-of-Fit 185

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

So the statistic A2 has the value 4 The probability under the nullhypothesis that the A2 statistic assumes a value gtfrac14 4 is 00286 So thenull hypothesis would be rejected The cumulative distribution functionfor this two-sample test statistic (with mfrac14 4 and nfrac14 6) looks like this

2 POWER OF THE A STATISTIC

In an article titled lsquolsquoEDF Statistics for Goodness-of-fit and SomeComparisonsrsquorsquo Stephens 1974 presented a chart comparing the powersof several different statistical tests The null hypothesis is that we havea uniform random number on the interval (0 1) There were sevenalternate distributions defined as follows

F FethxTHORN frac14 1 eth1 zTHORNk 0 ltfrac14 z ltfrac14 1 k frac14 15 2

G FethxTHORN frac14 2ethk1THORNzk 0 ltfrac14 z ltfrac14 5

FethxTHORN frac14 1 2ethk1THORNeth1 zTHORNk 5 ltfrac14 z ltfrac14 1 k frac14 15 2 3

H FethxTHORN frac14 05 2ethk1THORNeth5 zTHORNk 0 ltfrac14 z ltfrac14 5

FethxTHORN frac14 05thorn 2ethk1THORNethz 5THORNk 5 ltfrac14 z ltfrac14 1 k frac14 15 2

I computed the power of the A statistic in these same seven situationsThe results are shown in section four of this paper The power of theA statistic compares very favorably with both the Kolmogorov-SmirnovD statistic and the Cramer-von Mises W2 statistic

3 TABLES OF THE A STATISTIC

Tables of the exact distribution of A for sample sizes (n) rangingfrom 2 through 7 is shown in Table 3

The tables of the A statistic (Table 4) were computed using simula-tions The number of replications (N) is shown followed by the samplesize (n) and four critical values with the corresponding alpha levels

Value 0 1 2 3 4Cumulative density 1714 5143 8381 9714 1000

186 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

In each case the four alpha levels were selected to be near 020 010 005and 001

Here are some additional critical values for values of n from 8through 15 (Tables 5 and 6)

4 TABLES COMPARING POWER OF THE

A STATISTIC WITH OTHER

ONE-SAMPLE TESTS

As noted earlier the table of results comparing Kolmogorov-Smirnov D Cramer-von Mises W2 Kuiper V Watson U2 Anderson-Darling A2 Q (frac14Si ln zi) and Chi2 first appeared in an article byStephens (1974) in JASA I have simply added the column showing thepower of the A statistic as estimated by a number of simulations

Table 3

A nfrac14 2 nfrac14 3 nfrac14 4 nfrac14 5 nfrac14 6 nfrac14 7

0 0500 022222 009375 003840 001543 000611991 1000 066667 037500 019200 009259 004283932 092593 068750 044160 025720 013973773 100000 087500 067200 046296 029273524 096094 082560 064472 046256235 099219 091520 077761 061317996 100000 096352 086703 073131097 098656 092438 081845158 099616 095945 088053459 099936 097990 0923662710 100000 099096 0952885311 099640 0972077912 099880 0984233913 099970 0991597214 099996 0995813215 100000 0998074216 0999198617 0999708518 0999912619 0999980620 0999997621 10000000

A New One-Sample Test for Goodness-of-Fit 187

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

10000 8 9 02267 1000 18 32 019611 01183 38 010113 00543 44 004917 00089 57 0009

10000 9 11 02074 1000 19 33 020413 01176 40 010216 00462 48 004920 00106 63 0010

10000 10 13 02008 10000 20 37 0197316 00972 45 0099319 00406 53 0047724 00090 66 00092

10000 11 15 02018 1000 21 40 019719 00860 49 009521 00522 56 005127 00111 75 0009

10000 12 17 02015 1000 22 43 020020 01162 51 010024 00507 59 004930 00099 77 0010

10000 13 19 02056 1000 23 47 019823 01038 56 010027 00484 66 004935 00097 83 0010

10000 14 22 01909 1200 24 48 0200827 00945 58 0098331 00497 67 0049240 00085 87 00092

10000 15 23 02125 10000 25 51 0199129 00972 61 0103134 00468 72 0051543 00089 93 00107

1000 16 26 0192 1600 26 53 0203131 0092 65 0101336 0046 77 0048148 0010 97 00094

1000 17 29 0193 1600 27 57 0196934 0102 69 0098840 0045 80 0047551 0008 104 00100

188 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4 Continued

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

1600 28 59 01981 1600 38 96 02019

73 01019 118 0101386 00500 135 00500110 00100 181 00100

1600 29 66 01975 1600 39 99 0201380 01013 120 0100092 00481 140 00494117 00100 170 00094

10000 30 67 02029 10000 40 103 0201181 01025 126 0100095 00490 148 00496126 00095 191 00099

1600 31 71 01969 10000 50 145 0194587 00981 177 00984101 00500 205 00496128 00100 265 00099

1600 32 74 02038 10000 60 190 0199991 01000 232 00986107 00488 270 00498135 00100 348 00093

1600 33 77 01963 10000 70 240 0201990 01006 292 01027108 00500 340 00488141 00094 438 00095

1600 34 82 01975 10000 80 293 01988101 00988 357 01015117 00488 415 00530149 00100 535 00112

10000 35 84 02036 10000 90 349 01982102 01019 426 01015120 00493 495 00530152 00103 639 00112

1600 36 91 02019 10000 100 409 01930113 00981 499 00955132 00488 580 00467166 00100 748 00089

1600 37 90 01988110 01000128 00488172 00100

A New One-Sample Test for Goodness-of-Fit 189

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

5

nfrac148

nfrac149

nfrac1410

nfrac1411

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

902267

11

02074

13

02008

15

02018

10

01657

12

01580

14

01606

16

01628

11

01183

13

01176

15

01260

17

01336

12

00810

14

00874

16

00972

18

01091

13

00543

15

00633

17

00729

19

00860

14

00347

16

00462

18

00542

20

00670

15

00227

17

00335

19

00406

21

00522

16

00151

18

00241

20

00298

22

00417

17

00089

19

00157

21

00221

23

00334

20

00106

22

00166

24

00257

21

00062

23

00126

25

00194

24

00090

26

00147

27

00111

28

00082

190 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

6

nfrac1412

nfrac1413

nfrac1414

nfrac1415

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

17

02015

19

02056

21

02168

23

02125

18

01677

20

01733

22

01909

24

01873

19

01407

21

01488

23

01657

25

01648

20

01162

22

01249

24

01431

26

01454

21

00929

23

01038

25

01267

27

01285

22

00755

24

00878

26

01100

28

01132

23

00615

25

00736

27

00945

29

00972

24

00507

26

00606

28

00806

30

00841

25

00416

27

00484

29

00691

31

00723

26

00315

28

00410

30

00581

32

00624

27

00240

29

00334

31

00497

33

00542

28

00175

30

00278

32

00435

34

00468

29

00134

31

00221

33

00360

35

00403

30

00099

32

00184

34

00293

36

00344

33

00145

35

00236

37

00296

34

00120

36

00200

38

00240

35

00097

37

00175

39

00206

38

00141

40

00163

39

00119

41

00141

40

00085

42

00118

43

00089

A New One-Sample Test for Goodness-of-Fit 191

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

(Table 7) The alternative distribution functions are described in sectiontwo of this paper The sample sizes (n) were set equal to 10 20 and 40The alpha level was set at 010

REFERENCES

Bradley J V (1968) Distribution-Free Statistical Tests EnglewoodCliffs NJ Prentice-Hall

Gibbons J D (1976) Nonparametric Methods for Quantitative AnalysisNew York Holt Rinehart and Winston

Gibbons J D (1985) Nonparametric Statistical Inference New YorkMarcel Dekker

Table 7

Alternative D W2 V U2 A2 Q Chi2 A

(nfrac14 10 alphafrac14 010)Fkfrac1415 023 027 018 019 024 043 ndash 027Fkfrac1420 054 060 035 035 058 ndash ndash 058Gkfrac1415 009 007 022 023 006 ndash ndash 007Gkfrac1420 009 007 040 044 006 ndash ndash 010Gkfrac1430 021 021 081 086 018 ndash ndash 029Hkfrac1415 ndash ndash ndash ndash ndash ndash ndash 014Hkfrac1420 ndash ndash ndash ndash ndash ndash ndash 021

(nfrac14 20 alphafrac14 010)Fkfrac1415 038 046 025 028 046 068 ndash 046Fkfrac1420 078 087 061 060 087 097 059 088Gkfrac1415 013 011 032 034 010 011 ndash 010Gkfrac1420 025 025 071 077 028 025 ndash 029Gkfrac1430 063 079 099 099 084 ndash ndash 083Hkfrac1415 025 020 036 037 028 ndash ndash 017Hkfrac1420 047 044 071 077 054 ndash ndash 036

(nfrac14 40 alphafrac14 010)Fkfrac1415 060 070 043 043 ndash 089 040 073Fkfrac1420 098 099 091 089 ndash 100 089 099Gkfrac1415 019 022 057 061 ndash ndash 039 023Gkfrac1420 056 072 096 098 ndash ndash 085 075Gkfrac1430 ndash ndash ndash ndash ndash ndash ndash 100Hkfrac1415 036 032 058 063 ndash ndash ndash 027Hkfrac1420 071 080 096 098 ndash ndash ndash 075

192 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Hollander M Wolfe D A (1973) Nonparametric Statistical Methods2nd ed New York Wiley

Stephens M A (1974) EDF statistics for goodness-of-fit and somecomparisons (In theory and methods) J Amer Stat Assoc Theoryand Methods Section September 69(347)730ndash737

A New One-Sample Test for Goodness-of-Fit 193

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Request PermissionOrder Reprints

Reprints of this article can also be ordered at

httpwwwdekkercomservletproductDOI101081STA120026585

Request Permission or Order Reprints Instantly

Interested in copying and sharing this article In most cases US Copyright Law requires that you get permission from the articlersquos rightsholder before using copyrighted content

All information and materials found in this article including but not limited to text trademarks patents logos graphics and images (the Materials) are the copyrighted works and other forms of intellectual property of Marcel Dekker Inc or its licensors All rights not expressly granted are reserved

Get permission to lawfully reproduce and distribute the Materials or order reprints quickly and painlessly Simply click on the Request Permission Order Reprints link below and follow the instructions Visit the US Copyright Office for information on Fair Use limitations of US copyright law Please refer to The Association of American Publishersrsquo (AAP) website for guidelines on Fair Use in the Classroom

The Materials are for your personal use only and cannot be reformatted reposted resold or distributed by electronic means or otherwise without permission from Marcel Dekker Inc Marcel Dekker Inc grants you the limited right to display the Materials only on your personal computer or personal wireless device and to copy and download single copies of such Materials provided that any copyright trademark or other notice appearing on such Materials is also retained by displayed copied or downloaded as part of the Materials and is not removed or obscured and provided you do not edit modify alter or enhance the Materials Please refer to our Website User Agreement for more details

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Page 3: A New One-Sample Test for Goodness-of-Fit

ORDER REPRINTS

We can extend the definition of this statistic to a two-samplesituation

It is a test that provides excellent power My results show thatthe power is on a par with the Cramer-Von Mises one-sampletest for goodness-of-fit

This article contains five sections as follows

1 Defining the new statistic A

2 Description of the tests of power

3 Tables of the distribution for the A statistic

4 Table summarizing the results of the power tests

5 A brief bibliography

Key Words Distribution-free Goodness-of-fit Greatest integerfunction Non-parametric test One-sample test Two-sample test

1 DEFINITION OF THE A STATISTIC

Let Xi ifrac14 1 n denote n observations from a population PxThat is the Xi are independent and identically distributed We wish totest H0 Px is a population with some specified probability density func-tion (pdf )

Begin by partitioning the range of the pdf into n equal and non-over-lapping intervals (That is the integral of the pdf over each of these inter-valsfrac14 1=n) Under the null hypothesis we would expect one observationfrom our sample to occur in each of these n intervals (The expectedvalues of the order statistics are i=(nthorn 1) where i is an integer in therange from 1 to n inclusive) Our one-sample A statistic is a measureof how much the actual observations deviate from this expectation

If you think of this situation as analogous to a row of boxes each ofwhich contains from 0 to n balls with the total of all balls equal to thetotal number of boxes (n) then you may think of computing this statisticas equivalent to computing the minimum number of lsquolsquomovesrsquorsquo required toproduce a row of boxes with one and only one ball in each box And alsquolsquomoversquorsquo is defined as a move of one ball from one box to an adjacentbox An example may be useful

Suppose we have a random sample comprising the following sevenvalues 9088 9328 9351 1023 1044 1048 and 1053 We wish to testthe hypothesis that these seven values were drawn from a normal distri-bution with meanfrac14 100 and standard deviationfrac14 5 We compute the

182 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

z-scores and the corresponding values from the normal distribution(1824) 0034 (1344) 0089 (1298) 0097 (046) 0678 (088)0811 (096) 0832 and (106) 0855 We proceed as if we were testingthe null hypothesis that these seven decimal numbers (0034 through0855) came from a uniform distribution on the interval (000 100)We begin by defining seven intervals and determining the number ofobservations in each

(0000 0143) 3(0143 0286) 0(0286 0429) 0(0429 0571) 0(0571 0714) 1(0714 0857) 3(0857 1000) 0

We now determine the number of moves required to produce the lsquolsquogroundstatersquorsquo (ie one ball in each box)

1stmove

2ndmove

3rdmove

4thmove

5thmove

6thmove

(0000 0143) 3 2 2 1 1 1 1(0143 0286) 0 1 0 1 1 1 1(0286 0429) 0 0 1 1 1 1 1(0429 0571) 0 0 0 0 0 1 1(0571 0714) 1 1 1 1 2 1 1(0714 0857) 3 3 3 3 2 2 1(0857 1000) 0 0 0 0 0 0 1

So the computed value of the A statistic is 6 The probability under thenull hypothesis that the A statistic assumes a value gtfrac14 6 is 03868 Thisalpha level would generally not be considered significant and so the nullhypothesis would not be rejected

The mathematical description of the statistic is simply a summationof a number of terms We begin with a few definitions As already men-tioned the Xi ifrac14 1 n denote n observations from a population PxLet Si ifrac14 1 n denote the order statistics Let F(si) be the value of thecumulative distribution function for the ith order statistic Finally letGif( ) represent the greatest integer function We define A as follows

A frac14Xn

ifrac141

jGifethnFethsiTHORN thorn 1THORN ij

A New One-Sample Test for Goodness-of-Fit 183

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

So for our example

Gifeth7 0034thorn 1THORN frac14 1 j1 1j frac14 0

Gifeth7 0089thorn 1THORN frac14 1 j1 2j frac14 1

Gifeth7 0097thorn 1THORN frac14 1 j1 3j frac14 2

Gifeth7 0678thorn 1THORN frac14 5 j5 4j frac14 1

Gifeth7 0812thorn 1THORN frac14 6 j6 5j frac14 1

Gifeth7 0834thorn 1THORN frac14 6 j6 6j frac14 0

Gifeth7 0855thorn 1THORN frac14 6 j6 7j frac14 1

And the value of A is 0thorn 1thorn 2thorn 1thorn 1thorn 0thorn 1frac14 6 as before It should beobvious that the absolute values of the seven differences computed aboveactually correspond to the number of lsquolsquomovesrsquorsquo required to place each ofthe n order statistics into its corresponding lsquolsquoboxrsquorsquo Thus the third orderstatistic which lsquolsquooriginated in box 1rsquorsquo must be moved twice in order tolsquolsquoend up in box 3rsquorsquo

This statistic shares with the chi-square the concept of partitioningthe range of the probability distribution However it has the decidedadvantage of being useful in situations where we are able to obtain onlya small sample

In section three of this article I have published the exact distributionof the A statistic for values of nfrac14 2 3 4 5 6 and 7 For larger values of nI used computer simulations to generate the approximate distributions

After completing this work I was able to compute some formulaethat give excellent predictions of the critical values for large values of n(These formulae were computed by using stepwise regression techniques)

The resulting formulae (for alpha levelsfrac14 01 05 10 and 20) aregiven in the Table 1 The values computed from these formulae fornfrac14 20 30 40 50 60 70 80 90 and 100 are given in Table 2

Obviously these are good estimations of the values computedthrough the simulation and appearing in section three of this article

Another interesting fact is that we could easily extend the definitionof this statistic to test the two-sample problem in the following way

Let Xi ifrac14 1 m denote m observations from a population Pxand let F(x) be the cdf Let yj jfrac14 1 n denote n observations froma second population Py and let G(y) be the cdf We wish to test the nullhypothesis Ffrac14G against the broad alternative hypothesis F not frac14 GAssume that m is less than or equal to n Let tk kfrac14 1 (mthorn n)be the order statistics from the combined samples We construct thetwo-sample statistic A2 using the ranks (rirsquos) of the original xirsquos in the

184 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

combined sample

A2 frac14Xm

ifrac141

jGifethethriTHORN ethm=ethmthorn nthorn 1THORNTHORN thorn 1THORN ij

Example We have a sample of size 4 from the x population and asample of size 6 from the y population

The x observations (and ranks) are

089eth1THORN 097eth2THORN 125eth3THORN 498eth5THORN

The y observations (and ranks) are

134eth4THORN 678eth6THORN 714eth7THORN 834eth8THORN 912eth9THORN 934eth10THORN

We compute

Gifetheth1THORN eth4=11THORN thorn 1THORN frac14 1 j1 1j frac14 0

Gifetheth2THORN eth4=11THORN thorn 1THORN frac14 1 j1 2j frac14 1

Gifetheth3THORN eth4=11THORN thorn 1THORN frac14 2 j2 3j frac14 1

Gifetheth5THORN eth4=11THORN thorn 1THORN frac14 2 j2 4j frac14 2

Table 2

Values of n

Alpha levels 20 30 40 50 60 70 80 90 100

020 37 68 104 145 190 240 293 349 409010 45 82 127 177 232 292 357 426 499005 52 96 147 205 270 340 415 495 580001 67 123 190 265 348 438 535 639 748

Table 1

Alpha level Critical value

020 Gif(04081 n(3=2)thorn 1)010 Gif(04981 n(3=2)thorn 1)005 Gif(05797 n(3=2)thorn 1)001 Gif(07473 n(3=2)thorn 1)

A New One-Sample Test for Goodness-of-Fit 185

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

So the statistic A2 has the value 4 The probability under the nullhypothesis that the A2 statistic assumes a value gtfrac14 4 is 00286 So thenull hypothesis would be rejected The cumulative distribution functionfor this two-sample test statistic (with mfrac14 4 and nfrac14 6) looks like this

2 POWER OF THE A STATISTIC

In an article titled lsquolsquoEDF Statistics for Goodness-of-fit and SomeComparisonsrsquorsquo Stephens 1974 presented a chart comparing the powersof several different statistical tests The null hypothesis is that we havea uniform random number on the interval (0 1) There were sevenalternate distributions defined as follows

F FethxTHORN frac14 1 eth1 zTHORNk 0 ltfrac14 z ltfrac14 1 k frac14 15 2

G FethxTHORN frac14 2ethk1THORNzk 0 ltfrac14 z ltfrac14 5

FethxTHORN frac14 1 2ethk1THORNeth1 zTHORNk 5 ltfrac14 z ltfrac14 1 k frac14 15 2 3

H FethxTHORN frac14 05 2ethk1THORNeth5 zTHORNk 0 ltfrac14 z ltfrac14 5

FethxTHORN frac14 05thorn 2ethk1THORNethz 5THORNk 5 ltfrac14 z ltfrac14 1 k frac14 15 2

I computed the power of the A statistic in these same seven situationsThe results are shown in section four of this paper The power of theA statistic compares very favorably with both the Kolmogorov-SmirnovD statistic and the Cramer-von Mises W2 statistic

3 TABLES OF THE A STATISTIC

Tables of the exact distribution of A for sample sizes (n) rangingfrom 2 through 7 is shown in Table 3

The tables of the A statistic (Table 4) were computed using simula-tions The number of replications (N) is shown followed by the samplesize (n) and four critical values with the corresponding alpha levels

Value 0 1 2 3 4Cumulative density 1714 5143 8381 9714 1000

186 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

In each case the four alpha levels were selected to be near 020 010 005and 001

Here are some additional critical values for values of n from 8through 15 (Tables 5 and 6)

4 TABLES COMPARING POWER OF THE

A STATISTIC WITH OTHER

ONE-SAMPLE TESTS

As noted earlier the table of results comparing Kolmogorov-Smirnov D Cramer-von Mises W2 Kuiper V Watson U2 Anderson-Darling A2 Q (frac14Si ln zi) and Chi2 first appeared in an article byStephens (1974) in JASA I have simply added the column showing thepower of the A statistic as estimated by a number of simulations

Table 3

A nfrac14 2 nfrac14 3 nfrac14 4 nfrac14 5 nfrac14 6 nfrac14 7

0 0500 022222 009375 003840 001543 000611991 1000 066667 037500 019200 009259 004283932 092593 068750 044160 025720 013973773 100000 087500 067200 046296 029273524 096094 082560 064472 046256235 099219 091520 077761 061317996 100000 096352 086703 073131097 098656 092438 081845158 099616 095945 088053459 099936 097990 0923662710 100000 099096 0952885311 099640 0972077912 099880 0984233913 099970 0991597214 099996 0995813215 100000 0998074216 0999198617 0999708518 0999912619 0999980620 0999997621 10000000

A New One-Sample Test for Goodness-of-Fit 187

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

10000 8 9 02267 1000 18 32 019611 01183 38 010113 00543 44 004917 00089 57 0009

10000 9 11 02074 1000 19 33 020413 01176 40 010216 00462 48 004920 00106 63 0010

10000 10 13 02008 10000 20 37 0197316 00972 45 0099319 00406 53 0047724 00090 66 00092

10000 11 15 02018 1000 21 40 019719 00860 49 009521 00522 56 005127 00111 75 0009

10000 12 17 02015 1000 22 43 020020 01162 51 010024 00507 59 004930 00099 77 0010

10000 13 19 02056 1000 23 47 019823 01038 56 010027 00484 66 004935 00097 83 0010

10000 14 22 01909 1200 24 48 0200827 00945 58 0098331 00497 67 0049240 00085 87 00092

10000 15 23 02125 10000 25 51 0199129 00972 61 0103134 00468 72 0051543 00089 93 00107

1000 16 26 0192 1600 26 53 0203131 0092 65 0101336 0046 77 0048148 0010 97 00094

1000 17 29 0193 1600 27 57 0196934 0102 69 0098840 0045 80 0047551 0008 104 00100

188 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4 Continued

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

1600 28 59 01981 1600 38 96 02019

73 01019 118 0101386 00500 135 00500110 00100 181 00100

1600 29 66 01975 1600 39 99 0201380 01013 120 0100092 00481 140 00494117 00100 170 00094

10000 30 67 02029 10000 40 103 0201181 01025 126 0100095 00490 148 00496126 00095 191 00099

1600 31 71 01969 10000 50 145 0194587 00981 177 00984101 00500 205 00496128 00100 265 00099

1600 32 74 02038 10000 60 190 0199991 01000 232 00986107 00488 270 00498135 00100 348 00093

1600 33 77 01963 10000 70 240 0201990 01006 292 01027108 00500 340 00488141 00094 438 00095

1600 34 82 01975 10000 80 293 01988101 00988 357 01015117 00488 415 00530149 00100 535 00112

10000 35 84 02036 10000 90 349 01982102 01019 426 01015120 00493 495 00530152 00103 639 00112

1600 36 91 02019 10000 100 409 01930113 00981 499 00955132 00488 580 00467166 00100 748 00089

1600 37 90 01988110 01000128 00488172 00100

A New One-Sample Test for Goodness-of-Fit 189

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

5

nfrac148

nfrac149

nfrac1410

nfrac1411

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

902267

11

02074

13

02008

15

02018

10

01657

12

01580

14

01606

16

01628

11

01183

13

01176

15

01260

17

01336

12

00810

14

00874

16

00972

18

01091

13

00543

15

00633

17

00729

19

00860

14

00347

16

00462

18

00542

20

00670

15

00227

17

00335

19

00406

21

00522

16

00151

18

00241

20

00298

22

00417

17

00089

19

00157

21

00221

23

00334

20

00106

22

00166

24

00257

21

00062

23

00126

25

00194

24

00090

26

00147

27

00111

28

00082

190 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

6

nfrac1412

nfrac1413

nfrac1414

nfrac1415

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

17

02015

19

02056

21

02168

23

02125

18

01677

20

01733

22

01909

24

01873

19

01407

21

01488

23

01657

25

01648

20

01162

22

01249

24

01431

26

01454

21

00929

23

01038

25

01267

27

01285

22

00755

24

00878

26

01100

28

01132

23

00615

25

00736

27

00945

29

00972

24

00507

26

00606

28

00806

30

00841

25

00416

27

00484

29

00691

31

00723

26

00315

28

00410

30

00581

32

00624

27

00240

29

00334

31

00497

33

00542

28

00175

30

00278

32

00435

34

00468

29

00134

31

00221

33

00360

35

00403

30

00099

32

00184

34

00293

36

00344

33

00145

35

00236

37

00296

34

00120

36

00200

38

00240

35

00097

37

00175

39

00206

38

00141

40

00163

39

00119

41

00141

40

00085

42

00118

43

00089

A New One-Sample Test for Goodness-of-Fit 191

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

(Table 7) The alternative distribution functions are described in sectiontwo of this paper The sample sizes (n) were set equal to 10 20 and 40The alpha level was set at 010

REFERENCES

Bradley J V (1968) Distribution-Free Statistical Tests EnglewoodCliffs NJ Prentice-Hall

Gibbons J D (1976) Nonparametric Methods for Quantitative AnalysisNew York Holt Rinehart and Winston

Gibbons J D (1985) Nonparametric Statistical Inference New YorkMarcel Dekker

Table 7

Alternative D W2 V U2 A2 Q Chi2 A

(nfrac14 10 alphafrac14 010)Fkfrac1415 023 027 018 019 024 043 ndash 027Fkfrac1420 054 060 035 035 058 ndash ndash 058Gkfrac1415 009 007 022 023 006 ndash ndash 007Gkfrac1420 009 007 040 044 006 ndash ndash 010Gkfrac1430 021 021 081 086 018 ndash ndash 029Hkfrac1415 ndash ndash ndash ndash ndash ndash ndash 014Hkfrac1420 ndash ndash ndash ndash ndash ndash ndash 021

(nfrac14 20 alphafrac14 010)Fkfrac1415 038 046 025 028 046 068 ndash 046Fkfrac1420 078 087 061 060 087 097 059 088Gkfrac1415 013 011 032 034 010 011 ndash 010Gkfrac1420 025 025 071 077 028 025 ndash 029Gkfrac1430 063 079 099 099 084 ndash ndash 083Hkfrac1415 025 020 036 037 028 ndash ndash 017Hkfrac1420 047 044 071 077 054 ndash ndash 036

(nfrac14 40 alphafrac14 010)Fkfrac1415 060 070 043 043 ndash 089 040 073Fkfrac1420 098 099 091 089 ndash 100 089 099Gkfrac1415 019 022 057 061 ndash ndash 039 023Gkfrac1420 056 072 096 098 ndash ndash 085 075Gkfrac1430 ndash ndash ndash ndash ndash ndash ndash 100Hkfrac1415 036 032 058 063 ndash ndash ndash 027Hkfrac1420 071 080 096 098 ndash ndash ndash 075

192 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Hollander M Wolfe D A (1973) Nonparametric Statistical Methods2nd ed New York Wiley

Stephens M A (1974) EDF statistics for goodness-of-fit and somecomparisons (In theory and methods) J Amer Stat Assoc Theoryand Methods Section September 69(347)730ndash737

A New One-Sample Test for Goodness-of-Fit 193

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Request PermissionOrder Reprints

Reprints of this article can also be ordered at

httpwwwdekkercomservletproductDOI101081STA120026585

Request Permission or Order Reprints Instantly

Interested in copying and sharing this article In most cases US Copyright Law requires that you get permission from the articlersquos rightsholder before using copyrighted content

All information and materials found in this article including but not limited to text trademarks patents logos graphics and images (the Materials) are the copyrighted works and other forms of intellectual property of Marcel Dekker Inc or its licensors All rights not expressly granted are reserved

Get permission to lawfully reproduce and distribute the Materials or order reprints quickly and painlessly Simply click on the Request Permission Order Reprints link below and follow the instructions Visit the US Copyright Office for information on Fair Use limitations of US copyright law Please refer to The Association of American Publishersrsquo (AAP) website for guidelines on Fair Use in the Classroom

The Materials are for your personal use only and cannot be reformatted reposted resold or distributed by electronic means or otherwise without permission from Marcel Dekker Inc Marcel Dekker Inc grants you the limited right to display the Materials only on your personal computer or personal wireless device and to copy and download single copies of such Materials provided that any copyright trademark or other notice appearing on such Materials is also retained by displayed copied or downloaded as part of the Materials and is not removed or obscured and provided you do not edit modify alter or enhance the Materials Please refer to our Website User Agreement for more details

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Page 4: A New One-Sample Test for Goodness-of-Fit

ORDER REPRINTS

z-scores and the corresponding values from the normal distribution(1824) 0034 (1344) 0089 (1298) 0097 (046) 0678 (088)0811 (096) 0832 and (106) 0855 We proceed as if we were testingthe null hypothesis that these seven decimal numbers (0034 through0855) came from a uniform distribution on the interval (000 100)We begin by defining seven intervals and determining the number ofobservations in each

(0000 0143) 3(0143 0286) 0(0286 0429) 0(0429 0571) 0(0571 0714) 1(0714 0857) 3(0857 1000) 0

We now determine the number of moves required to produce the lsquolsquogroundstatersquorsquo (ie one ball in each box)

1stmove

2ndmove

3rdmove

4thmove

5thmove

6thmove

(0000 0143) 3 2 2 1 1 1 1(0143 0286) 0 1 0 1 1 1 1(0286 0429) 0 0 1 1 1 1 1(0429 0571) 0 0 0 0 0 1 1(0571 0714) 1 1 1 1 2 1 1(0714 0857) 3 3 3 3 2 2 1(0857 1000) 0 0 0 0 0 0 1

So the computed value of the A statistic is 6 The probability under thenull hypothesis that the A statistic assumes a value gtfrac14 6 is 03868 Thisalpha level would generally not be considered significant and so the nullhypothesis would not be rejected

The mathematical description of the statistic is simply a summationof a number of terms We begin with a few definitions As already men-tioned the Xi ifrac14 1 n denote n observations from a population PxLet Si ifrac14 1 n denote the order statistics Let F(si) be the value of thecumulative distribution function for the ith order statistic Finally letGif( ) represent the greatest integer function We define A as follows

A frac14Xn

ifrac141

jGifethnFethsiTHORN thorn 1THORN ij

A New One-Sample Test for Goodness-of-Fit 183

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

So for our example

Gifeth7 0034thorn 1THORN frac14 1 j1 1j frac14 0

Gifeth7 0089thorn 1THORN frac14 1 j1 2j frac14 1

Gifeth7 0097thorn 1THORN frac14 1 j1 3j frac14 2

Gifeth7 0678thorn 1THORN frac14 5 j5 4j frac14 1

Gifeth7 0812thorn 1THORN frac14 6 j6 5j frac14 1

Gifeth7 0834thorn 1THORN frac14 6 j6 6j frac14 0

Gifeth7 0855thorn 1THORN frac14 6 j6 7j frac14 1

And the value of A is 0thorn 1thorn 2thorn 1thorn 1thorn 0thorn 1frac14 6 as before It should beobvious that the absolute values of the seven differences computed aboveactually correspond to the number of lsquolsquomovesrsquorsquo required to place each ofthe n order statistics into its corresponding lsquolsquoboxrsquorsquo Thus the third orderstatistic which lsquolsquooriginated in box 1rsquorsquo must be moved twice in order tolsquolsquoend up in box 3rsquorsquo

This statistic shares with the chi-square the concept of partitioningthe range of the probability distribution However it has the decidedadvantage of being useful in situations where we are able to obtain onlya small sample

In section three of this article I have published the exact distributionof the A statistic for values of nfrac14 2 3 4 5 6 and 7 For larger values of nI used computer simulations to generate the approximate distributions

After completing this work I was able to compute some formulaethat give excellent predictions of the critical values for large values of n(These formulae were computed by using stepwise regression techniques)

The resulting formulae (for alpha levelsfrac14 01 05 10 and 20) aregiven in the Table 1 The values computed from these formulae fornfrac14 20 30 40 50 60 70 80 90 and 100 are given in Table 2

Obviously these are good estimations of the values computedthrough the simulation and appearing in section three of this article

Another interesting fact is that we could easily extend the definitionof this statistic to test the two-sample problem in the following way

Let Xi ifrac14 1 m denote m observations from a population Pxand let F(x) be the cdf Let yj jfrac14 1 n denote n observations froma second population Py and let G(y) be the cdf We wish to test the nullhypothesis Ffrac14G against the broad alternative hypothesis F not frac14 GAssume that m is less than or equal to n Let tk kfrac14 1 (mthorn n)be the order statistics from the combined samples We construct thetwo-sample statistic A2 using the ranks (rirsquos) of the original xirsquos in the

184 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

combined sample

A2 frac14Xm

ifrac141

jGifethethriTHORN ethm=ethmthorn nthorn 1THORNTHORN thorn 1THORN ij

Example We have a sample of size 4 from the x population and asample of size 6 from the y population

The x observations (and ranks) are

089eth1THORN 097eth2THORN 125eth3THORN 498eth5THORN

The y observations (and ranks) are

134eth4THORN 678eth6THORN 714eth7THORN 834eth8THORN 912eth9THORN 934eth10THORN

We compute

Gifetheth1THORN eth4=11THORN thorn 1THORN frac14 1 j1 1j frac14 0

Gifetheth2THORN eth4=11THORN thorn 1THORN frac14 1 j1 2j frac14 1

Gifetheth3THORN eth4=11THORN thorn 1THORN frac14 2 j2 3j frac14 1

Gifetheth5THORN eth4=11THORN thorn 1THORN frac14 2 j2 4j frac14 2

Table 2

Values of n

Alpha levels 20 30 40 50 60 70 80 90 100

020 37 68 104 145 190 240 293 349 409010 45 82 127 177 232 292 357 426 499005 52 96 147 205 270 340 415 495 580001 67 123 190 265 348 438 535 639 748

Table 1

Alpha level Critical value

020 Gif(04081 n(3=2)thorn 1)010 Gif(04981 n(3=2)thorn 1)005 Gif(05797 n(3=2)thorn 1)001 Gif(07473 n(3=2)thorn 1)

A New One-Sample Test for Goodness-of-Fit 185

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

So the statistic A2 has the value 4 The probability under the nullhypothesis that the A2 statistic assumes a value gtfrac14 4 is 00286 So thenull hypothesis would be rejected The cumulative distribution functionfor this two-sample test statistic (with mfrac14 4 and nfrac14 6) looks like this

2 POWER OF THE A STATISTIC

In an article titled lsquolsquoEDF Statistics for Goodness-of-fit and SomeComparisonsrsquorsquo Stephens 1974 presented a chart comparing the powersof several different statistical tests The null hypothesis is that we havea uniform random number on the interval (0 1) There were sevenalternate distributions defined as follows

F FethxTHORN frac14 1 eth1 zTHORNk 0 ltfrac14 z ltfrac14 1 k frac14 15 2

G FethxTHORN frac14 2ethk1THORNzk 0 ltfrac14 z ltfrac14 5

FethxTHORN frac14 1 2ethk1THORNeth1 zTHORNk 5 ltfrac14 z ltfrac14 1 k frac14 15 2 3

H FethxTHORN frac14 05 2ethk1THORNeth5 zTHORNk 0 ltfrac14 z ltfrac14 5

FethxTHORN frac14 05thorn 2ethk1THORNethz 5THORNk 5 ltfrac14 z ltfrac14 1 k frac14 15 2

I computed the power of the A statistic in these same seven situationsThe results are shown in section four of this paper The power of theA statistic compares very favorably with both the Kolmogorov-SmirnovD statistic and the Cramer-von Mises W2 statistic

3 TABLES OF THE A STATISTIC

Tables of the exact distribution of A for sample sizes (n) rangingfrom 2 through 7 is shown in Table 3

The tables of the A statistic (Table 4) were computed using simula-tions The number of replications (N) is shown followed by the samplesize (n) and four critical values with the corresponding alpha levels

Value 0 1 2 3 4Cumulative density 1714 5143 8381 9714 1000

186 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

In each case the four alpha levels were selected to be near 020 010 005and 001

Here are some additional critical values for values of n from 8through 15 (Tables 5 and 6)

4 TABLES COMPARING POWER OF THE

A STATISTIC WITH OTHER

ONE-SAMPLE TESTS

As noted earlier the table of results comparing Kolmogorov-Smirnov D Cramer-von Mises W2 Kuiper V Watson U2 Anderson-Darling A2 Q (frac14Si ln zi) and Chi2 first appeared in an article byStephens (1974) in JASA I have simply added the column showing thepower of the A statistic as estimated by a number of simulations

Table 3

A nfrac14 2 nfrac14 3 nfrac14 4 nfrac14 5 nfrac14 6 nfrac14 7

0 0500 022222 009375 003840 001543 000611991 1000 066667 037500 019200 009259 004283932 092593 068750 044160 025720 013973773 100000 087500 067200 046296 029273524 096094 082560 064472 046256235 099219 091520 077761 061317996 100000 096352 086703 073131097 098656 092438 081845158 099616 095945 088053459 099936 097990 0923662710 100000 099096 0952885311 099640 0972077912 099880 0984233913 099970 0991597214 099996 0995813215 100000 0998074216 0999198617 0999708518 0999912619 0999980620 0999997621 10000000

A New One-Sample Test for Goodness-of-Fit 187

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

10000 8 9 02267 1000 18 32 019611 01183 38 010113 00543 44 004917 00089 57 0009

10000 9 11 02074 1000 19 33 020413 01176 40 010216 00462 48 004920 00106 63 0010

10000 10 13 02008 10000 20 37 0197316 00972 45 0099319 00406 53 0047724 00090 66 00092

10000 11 15 02018 1000 21 40 019719 00860 49 009521 00522 56 005127 00111 75 0009

10000 12 17 02015 1000 22 43 020020 01162 51 010024 00507 59 004930 00099 77 0010

10000 13 19 02056 1000 23 47 019823 01038 56 010027 00484 66 004935 00097 83 0010

10000 14 22 01909 1200 24 48 0200827 00945 58 0098331 00497 67 0049240 00085 87 00092

10000 15 23 02125 10000 25 51 0199129 00972 61 0103134 00468 72 0051543 00089 93 00107

1000 16 26 0192 1600 26 53 0203131 0092 65 0101336 0046 77 0048148 0010 97 00094

1000 17 29 0193 1600 27 57 0196934 0102 69 0098840 0045 80 0047551 0008 104 00100

188 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4 Continued

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

1600 28 59 01981 1600 38 96 02019

73 01019 118 0101386 00500 135 00500110 00100 181 00100

1600 29 66 01975 1600 39 99 0201380 01013 120 0100092 00481 140 00494117 00100 170 00094

10000 30 67 02029 10000 40 103 0201181 01025 126 0100095 00490 148 00496126 00095 191 00099

1600 31 71 01969 10000 50 145 0194587 00981 177 00984101 00500 205 00496128 00100 265 00099

1600 32 74 02038 10000 60 190 0199991 01000 232 00986107 00488 270 00498135 00100 348 00093

1600 33 77 01963 10000 70 240 0201990 01006 292 01027108 00500 340 00488141 00094 438 00095

1600 34 82 01975 10000 80 293 01988101 00988 357 01015117 00488 415 00530149 00100 535 00112

10000 35 84 02036 10000 90 349 01982102 01019 426 01015120 00493 495 00530152 00103 639 00112

1600 36 91 02019 10000 100 409 01930113 00981 499 00955132 00488 580 00467166 00100 748 00089

1600 37 90 01988110 01000128 00488172 00100

A New One-Sample Test for Goodness-of-Fit 189

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

5

nfrac148

nfrac149

nfrac1410

nfrac1411

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

902267

11

02074

13

02008

15

02018

10

01657

12

01580

14

01606

16

01628

11

01183

13

01176

15

01260

17

01336

12

00810

14

00874

16

00972

18

01091

13

00543

15

00633

17

00729

19

00860

14

00347

16

00462

18

00542

20

00670

15

00227

17

00335

19

00406

21

00522

16

00151

18

00241

20

00298

22

00417

17

00089

19

00157

21

00221

23

00334

20

00106

22

00166

24

00257

21

00062

23

00126

25

00194

24

00090

26

00147

27

00111

28

00082

190 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

6

nfrac1412

nfrac1413

nfrac1414

nfrac1415

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

17

02015

19

02056

21

02168

23

02125

18

01677

20

01733

22

01909

24

01873

19

01407

21

01488

23

01657

25

01648

20

01162

22

01249

24

01431

26

01454

21

00929

23

01038

25

01267

27

01285

22

00755

24

00878

26

01100

28

01132

23

00615

25

00736

27

00945

29

00972

24

00507

26

00606

28

00806

30

00841

25

00416

27

00484

29

00691

31

00723

26

00315

28

00410

30

00581

32

00624

27

00240

29

00334

31

00497

33

00542

28

00175

30

00278

32

00435

34

00468

29

00134

31

00221

33

00360

35

00403

30

00099

32

00184

34

00293

36

00344

33

00145

35

00236

37

00296

34

00120

36

00200

38

00240

35

00097

37

00175

39

00206

38

00141

40

00163

39

00119

41

00141

40

00085

42

00118

43

00089

A New One-Sample Test for Goodness-of-Fit 191

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

(Table 7) The alternative distribution functions are described in sectiontwo of this paper The sample sizes (n) were set equal to 10 20 and 40The alpha level was set at 010

REFERENCES

Bradley J V (1968) Distribution-Free Statistical Tests EnglewoodCliffs NJ Prentice-Hall

Gibbons J D (1976) Nonparametric Methods for Quantitative AnalysisNew York Holt Rinehart and Winston

Gibbons J D (1985) Nonparametric Statistical Inference New YorkMarcel Dekker

Table 7

Alternative D W2 V U2 A2 Q Chi2 A

(nfrac14 10 alphafrac14 010)Fkfrac1415 023 027 018 019 024 043 ndash 027Fkfrac1420 054 060 035 035 058 ndash ndash 058Gkfrac1415 009 007 022 023 006 ndash ndash 007Gkfrac1420 009 007 040 044 006 ndash ndash 010Gkfrac1430 021 021 081 086 018 ndash ndash 029Hkfrac1415 ndash ndash ndash ndash ndash ndash ndash 014Hkfrac1420 ndash ndash ndash ndash ndash ndash ndash 021

(nfrac14 20 alphafrac14 010)Fkfrac1415 038 046 025 028 046 068 ndash 046Fkfrac1420 078 087 061 060 087 097 059 088Gkfrac1415 013 011 032 034 010 011 ndash 010Gkfrac1420 025 025 071 077 028 025 ndash 029Gkfrac1430 063 079 099 099 084 ndash ndash 083Hkfrac1415 025 020 036 037 028 ndash ndash 017Hkfrac1420 047 044 071 077 054 ndash ndash 036

(nfrac14 40 alphafrac14 010)Fkfrac1415 060 070 043 043 ndash 089 040 073Fkfrac1420 098 099 091 089 ndash 100 089 099Gkfrac1415 019 022 057 061 ndash ndash 039 023Gkfrac1420 056 072 096 098 ndash ndash 085 075Gkfrac1430 ndash ndash ndash ndash ndash ndash ndash 100Hkfrac1415 036 032 058 063 ndash ndash ndash 027Hkfrac1420 071 080 096 098 ndash ndash ndash 075

192 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Hollander M Wolfe D A (1973) Nonparametric Statistical Methods2nd ed New York Wiley

Stephens M A (1974) EDF statistics for goodness-of-fit and somecomparisons (In theory and methods) J Amer Stat Assoc Theoryand Methods Section September 69(347)730ndash737

A New One-Sample Test for Goodness-of-Fit 193

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Request PermissionOrder Reprints

Reprints of this article can also be ordered at

httpwwwdekkercomservletproductDOI101081STA120026585

Request Permission or Order Reprints Instantly

Interested in copying and sharing this article In most cases US Copyright Law requires that you get permission from the articlersquos rightsholder before using copyrighted content

All information and materials found in this article including but not limited to text trademarks patents logos graphics and images (the Materials) are the copyrighted works and other forms of intellectual property of Marcel Dekker Inc or its licensors All rights not expressly granted are reserved

Get permission to lawfully reproduce and distribute the Materials or order reprints quickly and painlessly Simply click on the Request Permission Order Reprints link below and follow the instructions Visit the US Copyright Office for information on Fair Use limitations of US copyright law Please refer to The Association of American Publishersrsquo (AAP) website for guidelines on Fair Use in the Classroom

The Materials are for your personal use only and cannot be reformatted reposted resold or distributed by electronic means or otherwise without permission from Marcel Dekker Inc Marcel Dekker Inc grants you the limited right to display the Materials only on your personal computer or personal wireless device and to copy and download single copies of such Materials provided that any copyright trademark or other notice appearing on such Materials is also retained by displayed copied or downloaded as part of the Materials and is not removed or obscured and provided you do not edit modify alter or enhance the Materials Please refer to our Website User Agreement for more details

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Page 5: A New One-Sample Test for Goodness-of-Fit

ORDER REPRINTS

So for our example

Gifeth7 0034thorn 1THORN frac14 1 j1 1j frac14 0

Gifeth7 0089thorn 1THORN frac14 1 j1 2j frac14 1

Gifeth7 0097thorn 1THORN frac14 1 j1 3j frac14 2

Gifeth7 0678thorn 1THORN frac14 5 j5 4j frac14 1

Gifeth7 0812thorn 1THORN frac14 6 j6 5j frac14 1

Gifeth7 0834thorn 1THORN frac14 6 j6 6j frac14 0

Gifeth7 0855thorn 1THORN frac14 6 j6 7j frac14 1

And the value of A is 0thorn 1thorn 2thorn 1thorn 1thorn 0thorn 1frac14 6 as before It should beobvious that the absolute values of the seven differences computed aboveactually correspond to the number of lsquolsquomovesrsquorsquo required to place each ofthe n order statistics into its corresponding lsquolsquoboxrsquorsquo Thus the third orderstatistic which lsquolsquooriginated in box 1rsquorsquo must be moved twice in order tolsquolsquoend up in box 3rsquorsquo

This statistic shares with the chi-square the concept of partitioningthe range of the probability distribution However it has the decidedadvantage of being useful in situations where we are able to obtain onlya small sample

In section three of this article I have published the exact distributionof the A statistic for values of nfrac14 2 3 4 5 6 and 7 For larger values of nI used computer simulations to generate the approximate distributions

After completing this work I was able to compute some formulaethat give excellent predictions of the critical values for large values of n(These formulae were computed by using stepwise regression techniques)

The resulting formulae (for alpha levelsfrac14 01 05 10 and 20) aregiven in the Table 1 The values computed from these formulae fornfrac14 20 30 40 50 60 70 80 90 and 100 are given in Table 2

Obviously these are good estimations of the values computedthrough the simulation and appearing in section three of this article

Another interesting fact is that we could easily extend the definitionof this statistic to test the two-sample problem in the following way

Let Xi ifrac14 1 m denote m observations from a population Pxand let F(x) be the cdf Let yj jfrac14 1 n denote n observations froma second population Py and let G(y) be the cdf We wish to test the nullhypothesis Ffrac14G against the broad alternative hypothesis F not frac14 GAssume that m is less than or equal to n Let tk kfrac14 1 (mthorn n)be the order statistics from the combined samples We construct thetwo-sample statistic A2 using the ranks (rirsquos) of the original xirsquos in the

184 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

combined sample

A2 frac14Xm

ifrac141

jGifethethriTHORN ethm=ethmthorn nthorn 1THORNTHORN thorn 1THORN ij

Example We have a sample of size 4 from the x population and asample of size 6 from the y population

The x observations (and ranks) are

089eth1THORN 097eth2THORN 125eth3THORN 498eth5THORN

The y observations (and ranks) are

134eth4THORN 678eth6THORN 714eth7THORN 834eth8THORN 912eth9THORN 934eth10THORN

We compute

Gifetheth1THORN eth4=11THORN thorn 1THORN frac14 1 j1 1j frac14 0

Gifetheth2THORN eth4=11THORN thorn 1THORN frac14 1 j1 2j frac14 1

Gifetheth3THORN eth4=11THORN thorn 1THORN frac14 2 j2 3j frac14 1

Gifetheth5THORN eth4=11THORN thorn 1THORN frac14 2 j2 4j frac14 2

Table 2

Values of n

Alpha levels 20 30 40 50 60 70 80 90 100

020 37 68 104 145 190 240 293 349 409010 45 82 127 177 232 292 357 426 499005 52 96 147 205 270 340 415 495 580001 67 123 190 265 348 438 535 639 748

Table 1

Alpha level Critical value

020 Gif(04081 n(3=2)thorn 1)010 Gif(04981 n(3=2)thorn 1)005 Gif(05797 n(3=2)thorn 1)001 Gif(07473 n(3=2)thorn 1)

A New One-Sample Test for Goodness-of-Fit 185

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

So the statistic A2 has the value 4 The probability under the nullhypothesis that the A2 statistic assumes a value gtfrac14 4 is 00286 So thenull hypothesis would be rejected The cumulative distribution functionfor this two-sample test statistic (with mfrac14 4 and nfrac14 6) looks like this

2 POWER OF THE A STATISTIC

In an article titled lsquolsquoEDF Statistics for Goodness-of-fit and SomeComparisonsrsquorsquo Stephens 1974 presented a chart comparing the powersof several different statistical tests The null hypothesis is that we havea uniform random number on the interval (0 1) There were sevenalternate distributions defined as follows

F FethxTHORN frac14 1 eth1 zTHORNk 0 ltfrac14 z ltfrac14 1 k frac14 15 2

G FethxTHORN frac14 2ethk1THORNzk 0 ltfrac14 z ltfrac14 5

FethxTHORN frac14 1 2ethk1THORNeth1 zTHORNk 5 ltfrac14 z ltfrac14 1 k frac14 15 2 3

H FethxTHORN frac14 05 2ethk1THORNeth5 zTHORNk 0 ltfrac14 z ltfrac14 5

FethxTHORN frac14 05thorn 2ethk1THORNethz 5THORNk 5 ltfrac14 z ltfrac14 1 k frac14 15 2

I computed the power of the A statistic in these same seven situationsThe results are shown in section four of this paper The power of theA statistic compares very favorably with both the Kolmogorov-SmirnovD statistic and the Cramer-von Mises W2 statistic

3 TABLES OF THE A STATISTIC

Tables of the exact distribution of A for sample sizes (n) rangingfrom 2 through 7 is shown in Table 3

The tables of the A statistic (Table 4) were computed using simula-tions The number of replications (N) is shown followed by the samplesize (n) and four critical values with the corresponding alpha levels

Value 0 1 2 3 4Cumulative density 1714 5143 8381 9714 1000

186 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

In each case the four alpha levels were selected to be near 020 010 005and 001

Here are some additional critical values for values of n from 8through 15 (Tables 5 and 6)

4 TABLES COMPARING POWER OF THE

A STATISTIC WITH OTHER

ONE-SAMPLE TESTS

As noted earlier the table of results comparing Kolmogorov-Smirnov D Cramer-von Mises W2 Kuiper V Watson U2 Anderson-Darling A2 Q (frac14Si ln zi) and Chi2 first appeared in an article byStephens (1974) in JASA I have simply added the column showing thepower of the A statistic as estimated by a number of simulations

Table 3

A nfrac14 2 nfrac14 3 nfrac14 4 nfrac14 5 nfrac14 6 nfrac14 7

0 0500 022222 009375 003840 001543 000611991 1000 066667 037500 019200 009259 004283932 092593 068750 044160 025720 013973773 100000 087500 067200 046296 029273524 096094 082560 064472 046256235 099219 091520 077761 061317996 100000 096352 086703 073131097 098656 092438 081845158 099616 095945 088053459 099936 097990 0923662710 100000 099096 0952885311 099640 0972077912 099880 0984233913 099970 0991597214 099996 0995813215 100000 0998074216 0999198617 0999708518 0999912619 0999980620 0999997621 10000000

A New One-Sample Test for Goodness-of-Fit 187

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

10000 8 9 02267 1000 18 32 019611 01183 38 010113 00543 44 004917 00089 57 0009

10000 9 11 02074 1000 19 33 020413 01176 40 010216 00462 48 004920 00106 63 0010

10000 10 13 02008 10000 20 37 0197316 00972 45 0099319 00406 53 0047724 00090 66 00092

10000 11 15 02018 1000 21 40 019719 00860 49 009521 00522 56 005127 00111 75 0009

10000 12 17 02015 1000 22 43 020020 01162 51 010024 00507 59 004930 00099 77 0010

10000 13 19 02056 1000 23 47 019823 01038 56 010027 00484 66 004935 00097 83 0010

10000 14 22 01909 1200 24 48 0200827 00945 58 0098331 00497 67 0049240 00085 87 00092

10000 15 23 02125 10000 25 51 0199129 00972 61 0103134 00468 72 0051543 00089 93 00107

1000 16 26 0192 1600 26 53 0203131 0092 65 0101336 0046 77 0048148 0010 97 00094

1000 17 29 0193 1600 27 57 0196934 0102 69 0098840 0045 80 0047551 0008 104 00100

188 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4 Continued

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

1600 28 59 01981 1600 38 96 02019

73 01019 118 0101386 00500 135 00500110 00100 181 00100

1600 29 66 01975 1600 39 99 0201380 01013 120 0100092 00481 140 00494117 00100 170 00094

10000 30 67 02029 10000 40 103 0201181 01025 126 0100095 00490 148 00496126 00095 191 00099

1600 31 71 01969 10000 50 145 0194587 00981 177 00984101 00500 205 00496128 00100 265 00099

1600 32 74 02038 10000 60 190 0199991 01000 232 00986107 00488 270 00498135 00100 348 00093

1600 33 77 01963 10000 70 240 0201990 01006 292 01027108 00500 340 00488141 00094 438 00095

1600 34 82 01975 10000 80 293 01988101 00988 357 01015117 00488 415 00530149 00100 535 00112

10000 35 84 02036 10000 90 349 01982102 01019 426 01015120 00493 495 00530152 00103 639 00112

1600 36 91 02019 10000 100 409 01930113 00981 499 00955132 00488 580 00467166 00100 748 00089

1600 37 90 01988110 01000128 00488172 00100

A New One-Sample Test for Goodness-of-Fit 189

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

5

nfrac148

nfrac149

nfrac1410

nfrac1411

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

902267

11

02074

13

02008

15

02018

10

01657

12

01580

14

01606

16

01628

11

01183

13

01176

15

01260

17

01336

12

00810

14

00874

16

00972

18

01091

13

00543

15

00633

17

00729

19

00860

14

00347

16

00462

18

00542

20

00670

15

00227

17

00335

19

00406

21

00522

16

00151

18

00241

20

00298

22

00417

17

00089

19

00157

21

00221

23

00334

20

00106

22

00166

24

00257

21

00062

23

00126

25

00194

24

00090

26

00147

27

00111

28

00082

190 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

6

nfrac1412

nfrac1413

nfrac1414

nfrac1415

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

17

02015

19

02056

21

02168

23

02125

18

01677

20

01733

22

01909

24

01873

19

01407

21

01488

23

01657

25

01648

20

01162

22

01249

24

01431

26

01454

21

00929

23

01038

25

01267

27

01285

22

00755

24

00878

26

01100

28

01132

23

00615

25

00736

27

00945

29

00972

24

00507

26

00606

28

00806

30

00841

25

00416

27

00484

29

00691

31

00723

26

00315

28

00410

30

00581

32

00624

27

00240

29

00334

31

00497

33

00542

28

00175

30

00278

32

00435

34

00468

29

00134

31

00221

33

00360

35

00403

30

00099

32

00184

34

00293

36

00344

33

00145

35

00236

37

00296

34

00120

36

00200

38

00240

35

00097

37

00175

39

00206

38

00141

40

00163

39

00119

41

00141

40

00085

42

00118

43

00089

A New One-Sample Test for Goodness-of-Fit 191

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

(Table 7) The alternative distribution functions are described in sectiontwo of this paper The sample sizes (n) were set equal to 10 20 and 40The alpha level was set at 010

REFERENCES

Bradley J V (1968) Distribution-Free Statistical Tests EnglewoodCliffs NJ Prentice-Hall

Gibbons J D (1976) Nonparametric Methods for Quantitative AnalysisNew York Holt Rinehart and Winston

Gibbons J D (1985) Nonparametric Statistical Inference New YorkMarcel Dekker

Table 7

Alternative D W2 V U2 A2 Q Chi2 A

(nfrac14 10 alphafrac14 010)Fkfrac1415 023 027 018 019 024 043 ndash 027Fkfrac1420 054 060 035 035 058 ndash ndash 058Gkfrac1415 009 007 022 023 006 ndash ndash 007Gkfrac1420 009 007 040 044 006 ndash ndash 010Gkfrac1430 021 021 081 086 018 ndash ndash 029Hkfrac1415 ndash ndash ndash ndash ndash ndash ndash 014Hkfrac1420 ndash ndash ndash ndash ndash ndash ndash 021

(nfrac14 20 alphafrac14 010)Fkfrac1415 038 046 025 028 046 068 ndash 046Fkfrac1420 078 087 061 060 087 097 059 088Gkfrac1415 013 011 032 034 010 011 ndash 010Gkfrac1420 025 025 071 077 028 025 ndash 029Gkfrac1430 063 079 099 099 084 ndash ndash 083Hkfrac1415 025 020 036 037 028 ndash ndash 017Hkfrac1420 047 044 071 077 054 ndash ndash 036

(nfrac14 40 alphafrac14 010)Fkfrac1415 060 070 043 043 ndash 089 040 073Fkfrac1420 098 099 091 089 ndash 100 089 099Gkfrac1415 019 022 057 061 ndash ndash 039 023Gkfrac1420 056 072 096 098 ndash ndash 085 075Gkfrac1430 ndash ndash ndash ndash ndash ndash ndash 100Hkfrac1415 036 032 058 063 ndash ndash ndash 027Hkfrac1420 071 080 096 098 ndash ndash ndash 075

192 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Hollander M Wolfe D A (1973) Nonparametric Statistical Methods2nd ed New York Wiley

Stephens M A (1974) EDF statistics for goodness-of-fit and somecomparisons (In theory and methods) J Amer Stat Assoc Theoryand Methods Section September 69(347)730ndash737

A New One-Sample Test for Goodness-of-Fit 193

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Request PermissionOrder Reprints

Reprints of this article can also be ordered at

httpwwwdekkercomservletproductDOI101081STA120026585

Request Permission or Order Reprints Instantly

Interested in copying and sharing this article In most cases US Copyright Law requires that you get permission from the articlersquos rightsholder before using copyrighted content

All information and materials found in this article including but not limited to text trademarks patents logos graphics and images (the Materials) are the copyrighted works and other forms of intellectual property of Marcel Dekker Inc or its licensors All rights not expressly granted are reserved

Get permission to lawfully reproduce and distribute the Materials or order reprints quickly and painlessly Simply click on the Request Permission Order Reprints link below and follow the instructions Visit the US Copyright Office for information on Fair Use limitations of US copyright law Please refer to The Association of American Publishersrsquo (AAP) website for guidelines on Fair Use in the Classroom

The Materials are for your personal use only and cannot be reformatted reposted resold or distributed by electronic means or otherwise without permission from Marcel Dekker Inc Marcel Dekker Inc grants you the limited right to display the Materials only on your personal computer or personal wireless device and to copy and download single copies of such Materials provided that any copyright trademark or other notice appearing on such Materials is also retained by displayed copied or downloaded as part of the Materials and is not removed or obscured and provided you do not edit modify alter or enhance the Materials Please refer to our Website User Agreement for more details

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Page 6: A New One-Sample Test for Goodness-of-Fit

ORDER REPRINTS

combined sample

A2 frac14Xm

ifrac141

jGifethethriTHORN ethm=ethmthorn nthorn 1THORNTHORN thorn 1THORN ij

Example We have a sample of size 4 from the x population and asample of size 6 from the y population

The x observations (and ranks) are

089eth1THORN 097eth2THORN 125eth3THORN 498eth5THORN

The y observations (and ranks) are

134eth4THORN 678eth6THORN 714eth7THORN 834eth8THORN 912eth9THORN 934eth10THORN

We compute

Gifetheth1THORN eth4=11THORN thorn 1THORN frac14 1 j1 1j frac14 0

Gifetheth2THORN eth4=11THORN thorn 1THORN frac14 1 j1 2j frac14 1

Gifetheth3THORN eth4=11THORN thorn 1THORN frac14 2 j2 3j frac14 1

Gifetheth5THORN eth4=11THORN thorn 1THORN frac14 2 j2 4j frac14 2

Table 2

Values of n

Alpha levels 20 30 40 50 60 70 80 90 100

020 37 68 104 145 190 240 293 349 409010 45 82 127 177 232 292 357 426 499005 52 96 147 205 270 340 415 495 580001 67 123 190 265 348 438 535 639 748

Table 1

Alpha level Critical value

020 Gif(04081 n(3=2)thorn 1)010 Gif(04981 n(3=2)thorn 1)005 Gif(05797 n(3=2)thorn 1)001 Gif(07473 n(3=2)thorn 1)

A New One-Sample Test for Goodness-of-Fit 185

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

So the statistic A2 has the value 4 The probability under the nullhypothesis that the A2 statistic assumes a value gtfrac14 4 is 00286 So thenull hypothesis would be rejected The cumulative distribution functionfor this two-sample test statistic (with mfrac14 4 and nfrac14 6) looks like this

2 POWER OF THE A STATISTIC

In an article titled lsquolsquoEDF Statistics for Goodness-of-fit and SomeComparisonsrsquorsquo Stephens 1974 presented a chart comparing the powersof several different statistical tests The null hypothesis is that we havea uniform random number on the interval (0 1) There were sevenalternate distributions defined as follows

F FethxTHORN frac14 1 eth1 zTHORNk 0 ltfrac14 z ltfrac14 1 k frac14 15 2

G FethxTHORN frac14 2ethk1THORNzk 0 ltfrac14 z ltfrac14 5

FethxTHORN frac14 1 2ethk1THORNeth1 zTHORNk 5 ltfrac14 z ltfrac14 1 k frac14 15 2 3

H FethxTHORN frac14 05 2ethk1THORNeth5 zTHORNk 0 ltfrac14 z ltfrac14 5

FethxTHORN frac14 05thorn 2ethk1THORNethz 5THORNk 5 ltfrac14 z ltfrac14 1 k frac14 15 2

I computed the power of the A statistic in these same seven situationsThe results are shown in section four of this paper The power of theA statistic compares very favorably with both the Kolmogorov-SmirnovD statistic and the Cramer-von Mises W2 statistic

3 TABLES OF THE A STATISTIC

Tables of the exact distribution of A for sample sizes (n) rangingfrom 2 through 7 is shown in Table 3

The tables of the A statistic (Table 4) were computed using simula-tions The number of replications (N) is shown followed by the samplesize (n) and four critical values with the corresponding alpha levels

Value 0 1 2 3 4Cumulative density 1714 5143 8381 9714 1000

186 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

In each case the four alpha levels were selected to be near 020 010 005and 001

Here are some additional critical values for values of n from 8through 15 (Tables 5 and 6)

4 TABLES COMPARING POWER OF THE

A STATISTIC WITH OTHER

ONE-SAMPLE TESTS

As noted earlier the table of results comparing Kolmogorov-Smirnov D Cramer-von Mises W2 Kuiper V Watson U2 Anderson-Darling A2 Q (frac14Si ln zi) and Chi2 first appeared in an article byStephens (1974) in JASA I have simply added the column showing thepower of the A statistic as estimated by a number of simulations

Table 3

A nfrac14 2 nfrac14 3 nfrac14 4 nfrac14 5 nfrac14 6 nfrac14 7

0 0500 022222 009375 003840 001543 000611991 1000 066667 037500 019200 009259 004283932 092593 068750 044160 025720 013973773 100000 087500 067200 046296 029273524 096094 082560 064472 046256235 099219 091520 077761 061317996 100000 096352 086703 073131097 098656 092438 081845158 099616 095945 088053459 099936 097990 0923662710 100000 099096 0952885311 099640 0972077912 099880 0984233913 099970 0991597214 099996 0995813215 100000 0998074216 0999198617 0999708518 0999912619 0999980620 0999997621 10000000

A New One-Sample Test for Goodness-of-Fit 187

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

10000 8 9 02267 1000 18 32 019611 01183 38 010113 00543 44 004917 00089 57 0009

10000 9 11 02074 1000 19 33 020413 01176 40 010216 00462 48 004920 00106 63 0010

10000 10 13 02008 10000 20 37 0197316 00972 45 0099319 00406 53 0047724 00090 66 00092

10000 11 15 02018 1000 21 40 019719 00860 49 009521 00522 56 005127 00111 75 0009

10000 12 17 02015 1000 22 43 020020 01162 51 010024 00507 59 004930 00099 77 0010

10000 13 19 02056 1000 23 47 019823 01038 56 010027 00484 66 004935 00097 83 0010

10000 14 22 01909 1200 24 48 0200827 00945 58 0098331 00497 67 0049240 00085 87 00092

10000 15 23 02125 10000 25 51 0199129 00972 61 0103134 00468 72 0051543 00089 93 00107

1000 16 26 0192 1600 26 53 0203131 0092 65 0101336 0046 77 0048148 0010 97 00094

1000 17 29 0193 1600 27 57 0196934 0102 69 0098840 0045 80 0047551 0008 104 00100

188 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4 Continued

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

1600 28 59 01981 1600 38 96 02019

73 01019 118 0101386 00500 135 00500110 00100 181 00100

1600 29 66 01975 1600 39 99 0201380 01013 120 0100092 00481 140 00494117 00100 170 00094

10000 30 67 02029 10000 40 103 0201181 01025 126 0100095 00490 148 00496126 00095 191 00099

1600 31 71 01969 10000 50 145 0194587 00981 177 00984101 00500 205 00496128 00100 265 00099

1600 32 74 02038 10000 60 190 0199991 01000 232 00986107 00488 270 00498135 00100 348 00093

1600 33 77 01963 10000 70 240 0201990 01006 292 01027108 00500 340 00488141 00094 438 00095

1600 34 82 01975 10000 80 293 01988101 00988 357 01015117 00488 415 00530149 00100 535 00112

10000 35 84 02036 10000 90 349 01982102 01019 426 01015120 00493 495 00530152 00103 639 00112

1600 36 91 02019 10000 100 409 01930113 00981 499 00955132 00488 580 00467166 00100 748 00089

1600 37 90 01988110 01000128 00488172 00100

A New One-Sample Test for Goodness-of-Fit 189

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

5

nfrac148

nfrac149

nfrac1410

nfrac1411

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

902267

11

02074

13

02008

15

02018

10

01657

12

01580

14

01606

16

01628

11

01183

13

01176

15

01260

17

01336

12

00810

14

00874

16

00972

18

01091

13

00543

15

00633

17

00729

19

00860

14

00347

16

00462

18

00542

20

00670

15

00227

17

00335

19

00406

21

00522

16

00151

18

00241

20

00298

22

00417

17

00089

19

00157

21

00221

23

00334

20

00106

22

00166

24

00257

21

00062

23

00126

25

00194

24

00090

26

00147

27

00111

28

00082

190 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

6

nfrac1412

nfrac1413

nfrac1414

nfrac1415

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

17

02015

19

02056

21

02168

23

02125

18

01677

20

01733

22

01909

24

01873

19

01407

21

01488

23

01657

25

01648

20

01162

22

01249

24

01431

26

01454

21

00929

23

01038

25

01267

27

01285

22

00755

24

00878

26

01100

28

01132

23

00615

25

00736

27

00945

29

00972

24

00507

26

00606

28

00806

30

00841

25

00416

27

00484

29

00691

31

00723

26

00315

28

00410

30

00581

32

00624

27

00240

29

00334

31

00497

33

00542

28

00175

30

00278

32

00435

34

00468

29

00134

31

00221

33

00360

35

00403

30

00099

32

00184

34

00293

36

00344

33

00145

35

00236

37

00296

34

00120

36

00200

38

00240

35

00097

37

00175

39

00206

38

00141

40

00163

39

00119

41

00141

40

00085

42

00118

43

00089

A New One-Sample Test for Goodness-of-Fit 191

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

(Table 7) The alternative distribution functions are described in sectiontwo of this paper The sample sizes (n) were set equal to 10 20 and 40The alpha level was set at 010

REFERENCES

Bradley J V (1968) Distribution-Free Statistical Tests EnglewoodCliffs NJ Prentice-Hall

Gibbons J D (1976) Nonparametric Methods for Quantitative AnalysisNew York Holt Rinehart and Winston

Gibbons J D (1985) Nonparametric Statistical Inference New YorkMarcel Dekker

Table 7

Alternative D W2 V U2 A2 Q Chi2 A

(nfrac14 10 alphafrac14 010)Fkfrac1415 023 027 018 019 024 043 ndash 027Fkfrac1420 054 060 035 035 058 ndash ndash 058Gkfrac1415 009 007 022 023 006 ndash ndash 007Gkfrac1420 009 007 040 044 006 ndash ndash 010Gkfrac1430 021 021 081 086 018 ndash ndash 029Hkfrac1415 ndash ndash ndash ndash ndash ndash ndash 014Hkfrac1420 ndash ndash ndash ndash ndash ndash ndash 021

(nfrac14 20 alphafrac14 010)Fkfrac1415 038 046 025 028 046 068 ndash 046Fkfrac1420 078 087 061 060 087 097 059 088Gkfrac1415 013 011 032 034 010 011 ndash 010Gkfrac1420 025 025 071 077 028 025 ndash 029Gkfrac1430 063 079 099 099 084 ndash ndash 083Hkfrac1415 025 020 036 037 028 ndash ndash 017Hkfrac1420 047 044 071 077 054 ndash ndash 036

(nfrac14 40 alphafrac14 010)Fkfrac1415 060 070 043 043 ndash 089 040 073Fkfrac1420 098 099 091 089 ndash 100 089 099Gkfrac1415 019 022 057 061 ndash ndash 039 023Gkfrac1420 056 072 096 098 ndash ndash 085 075Gkfrac1430 ndash ndash ndash ndash ndash ndash ndash 100Hkfrac1415 036 032 058 063 ndash ndash ndash 027Hkfrac1420 071 080 096 098 ndash ndash ndash 075

192 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Hollander M Wolfe D A (1973) Nonparametric Statistical Methods2nd ed New York Wiley

Stephens M A (1974) EDF statistics for goodness-of-fit and somecomparisons (In theory and methods) J Amer Stat Assoc Theoryand Methods Section September 69(347)730ndash737

A New One-Sample Test for Goodness-of-Fit 193

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Request PermissionOrder Reprints

Reprints of this article can also be ordered at

httpwwwdekkercomservletproductDOI101081STA120026585

Request Permission or Order Reprints Instantly

Interested in copying and sharing this article In most cases US Copyright Law requires that you get permission from the articlersquos rightsholder before using copyrighted content

All information and materials found in this article including but not limited to text trademarks patents logos graphics and images (the Materials) are the copyrighted works and other forms of intellectual property of Marcel Dekker Inc or its licensors All rights not expressly granted are reserved

Get permission to lawfully reproduce and distribute the Materials or order reprints quickly and painlessly Simply click on the Request Permission Order Reprints link below and follow the instructions Visit the US Copyright Office for information on Fair Use limitations of US copyright law Please refer to The Association of American Publishersrsquo (AAP) website for guidelines on Fair Use in the Classroom

The Materials are for your personal use only and cannot be reformatted reposted resold or distributed by electronic means or otherwise without permission from Marcel Dekker Inc Marcel Dekker Inc grants you the limited right to display the Materials only on your personal computer or personal wireless device and to copy and download single copies of such Materials provided that any copyright trademark or other notice appearing on such Materials is also retained by displayed copied or downloaded as part of the Materials and is not removed or obscured and provided you do not edit modify alter or enhance the Materials Please refer to our Website User Agreement for more details

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Page 7: A New One-Sample Test for Goodness-of-Fit

ORDER REPRINTS

So the statistic A2 has the value 4 The probability under the nullhypothesis that the A2 statistic assumes a value gtfrac14 4 is 00286 So thenull hypothesis would be rejected The cumulative distribution functionfor this two-sample test statistic (with mfrac14 4 and nfrac14 6) looks like this

2 POWER OF THE A STATISTIC

In an article titled lsquolsquoEDF Statistics for Goodness-of-fit and SomeComparisonsrsquorsquo Stephens 1974 presented a chart comparing the powersof several different statistical tests The null hypothesis is that we havea uniform random number on the interval (0 1) There were sevenalternate distributions defined as follows

F FethxTHORN frac14 1 eth1 zTHORNk 0 ltfrac14 z ltfrac14 1 k frac14 15 2

G FethxTHORN frac14 2ethk1THORNzk 0 ltfrac14 z ltfrac14 5

FethxTHORN frac14 1 2ethk1THORNeth1 zTHORNk 5 ltfrac14 z ltfrac14 1 k frac14 15 2 3

H FethxTHORN frac14 05 2ethk1THORNeth5 zTHORNk 0 ltfrac14 z ltfrac14 5

FethxTHORN frac14 05thorn 2ethk1THORNethz 5THORNk 5 ltfrac14 z ltfrac14 1 k frac14 15 2

I computed the power of the A statistic in these same seven situationsThe results are shown in section four of this paper The power of theA statistic compares very favorably with both the Kolmogorov-SmirnovD statistic and the Cramer-von Mises W2 statistic

3 TABLES OF THE A STATISTIC

Tables of the exact distribution of A for sample sizes (n) rangingfrom 2 through 7 is shown in Table 3

The tables of the A statistic (Table 4) were computed using simula-tions The number of replications (N) is shown followed by the samplesize (n) and four critical values with the corresponding alpha levels

Value 0 1 2 3 4Cumulative density 1714 5143 8381 9714 1000

186 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

In each case the four alpha levels were selected to be near 020 010 005and 001

Here are some additional critical values for values of n from 8through 15 (Tables 5 and 6)

4 TABLES COMPARING POWER OF THE

A STATISTIC WITH OTHER

ONE-SAMPLE TESTS

As noted earlier the table of results comparing Kolmogorov-Smirnov D Cramer-von Mises W2 Kuiper V Watson U2 Anderson-Darling A2 Q (frac14Si ln zi) and Chi2 first appeared in an article byStephens (1974) in JASA I have simply added the column showing thepower of the A statistic as estimated by a number of simulations

Table 3

A nfrac14 2 nfrac14 3 nfrac14 4 nfrac14 5 nfrac14 6 nfrac14 7

0 0500 022222 009375 003840 001543 000611991 1000 066667 037500 019200 009259 004283932 092593 068750 044160 025720 013973773 100000 087500 067200 046296 029273524 096094 082560 064472 046256235 099219 091520 077761 061317996 100000 096352 086703 073131097 098656 092438 081845158 099616 095945 088053459 099936 097990 0923662710 100000 099096 0952885311 099640 0972077912 099880 0984233913 099970 0991597214 099996 0995813215 100000 0998074216 0999198617 0999708518 0999912619 0999980620 0999997621 10000000

A New One-Sample Test for Goodness-of-Fit 187

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

10000 8 9 02267 1000 18 32 019611 01183 38 010113 00543 44 004917 00089 57 0009

10000 9 11 02074 1000 19 33 020413 01176 40 010216 00462 48 004920 00106 63 0010

10000 10 13 02008 10000 20 37 0197316 00972 45 0099319 00406 53 0047724 00090 66 00092

10000 11 15 02018 1000 21 40 019719 00860 49 009521 00522 56 005127 00111 75 0009

10000 12 17 02015 1000 22 43 020020 01162 51 010024 00507 59 004930 00099 77 0010

10000 13 19 02056 1000 23 47 019823 01038 56 010027 00484 66 004935 00097 83 0010

10000 14 22 01909 1200 24 48 0200827 00945 58 0098331 00497 67 0049240 00085 87 00092

10000 15 23 02125 10000 25 51 0199129 00972 61 0103134 00468 72 0051543 00089 93 00107

1000 16 26 0192 1600 26 53 0203131 0092 65 0101336 0046 77 0048148 0010 97 00094

1000 17 29 0193 1600 27 57 0196934 0102 69 0098840 0045 80 0047551 0008 104 00100

188 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4 Continued

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

1600 28 59 01981 1600 38 96 02019

73 01019 118 0101386 00500 135 00500110 00100 181 00100

1600 29 66 01975 1600 39 99 0201380 01013 120 0100092 00481 140 00494117 00100 170 00094

10000 30 67 02029 10000 40 103 0201181 01025 126 0100095 00490 148 00496126 00095 191 00099

1600 31 71 01969 10000 50 145 0194587 00981 177 00984101 00500 205 00496128 00100 265 00099

1600 32 74 02038 10000 60 190 0199991 01000 232 00986107 00488 270 00498135 00100 348 00093

1600 33 77 01963 10000 70 240 0201990 01006 292 01027108 00500 340 00488141 00094 438 00095

1600 34 82 01975 10000 80 293 01988101 00988 357 01015117 00488 415 00530149 00100 535 00112

10000 35 84 02036 10000 90 349 01982102 01019 426 01015120 00493 495 00530152 00103 639 00112

1600 36 91 02019 10000 100 409 01930113 00981 499 00955132 00488 580 00467166 00100 748 00089

1600 37 90 01988110 01000128 00488172 00100

A New One-Sample Test for Goodness-of-Fit 189

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

5

nfrac148

nfrac149

nfrac1410

nfrac1411

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

902267

11

02074

13

02008

15

02018

10

01657

12

01580

14

01606

16

01628

11

01183

13

01176

15

01260

17

01336

12

00810

14

00874

16

00972

18

01091

13

00543

15

00633

17

00729

19

00860

14

00347

16

00462

18

00542

20

00670

15

00227

17

00335

19

00406

21

00522

16

00151

18

00241

20

00298

22

00417

17

00089

19

00157

21

00221

23

00334

20

00106

22

00166

24

00257

21

00062

23

00126

25

00194

24

00090

26

00147

27

00111

28

00082

190 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

6

nfrac1412

nfrac1413

nfrac1414

nfrac1415

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

17

02015

19

02056

21

02168

23

02125

18

01677

20

01733

22

01909

24

01873

19

01407

21

01488

23

01657

25

01648

20

01162

22

01249

24

01431

26

01454

21

00929

23

01038

25

01267

27

01285

22

00755

24

00878

26

01100

28

01132

23

00615

25

00736

27

00945

29

00972

24

00507

26

00606

28

00806

30

00841

25

00416

27

00484

29

00691

31

00723

26

00315

28

00410

30

00581

32

00624

27

00240

29

00334

31

00497

33

00542

28

00175

30

00278

32

00435

34

00468

29

00134

31

00221

33

00360

35

00403

30

00099

32

00184

34

00293

36

00344

33

00145

35

00236

37

00296

34

00120

36

00200

38

00240

35

00097

37

00175

39

00206

38

00141

40

00163

39

00119

41

00141

40

00085

42

00118

43

00089

A New One-Sample Test for Goodness-of-Fit 191

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

(Table 7) The alternative distribution functions are described in sectiontwo of this paper The sample sizes (n) were set equal to 10 20 and 40The alpha level was set at 010

REFERENCES

Bradley J V (1968) Distribution-Free Statistical Tests EnglewoodCliffs NJ Prentice-Hall

Gibbons J D (1976) Nonparametric Methods for Quantitative AnalysisNew York Holt Rinehart and Winston

Gibbons J D (1985) Nonparametric Statistical Inference New YorkMarcel Dekker

Table 7

Alternative D W2 V U2 A2 Q Chi2 A

(nfrac14 10 alphafrac14 010)Fkfrac1415 023 027 018 019 024 043 ndash 027Fkfrac1420 054 060 035 035 058 ndash ndash 058Gkfrac1415 009 007 022 023 006 ndash ndash 007Gkfrac1420 009 007 040 044 006 ndash ndash 010Gkfrac1430 021 021 081 086 018 ndash ndash 029Hkfrac1415 ndash ndash ndash ndash ndash ndash ndash 014Hkfrac1420 ndash ndash ndash ndash ndash ndash ndash 021

(nfrac14 20 alphafrac14 010)Fkfrac1415 038 046 025 028 046 068 ndash 046Fkfrac1420 078 087 061 060 087 097 059 088Gkfrac1415 013 011 032 034 010 011 ndash 010Gkfrac1420 025 025 071 077 028 025 ndash 029Gkfrac1430 063 079 099 099 084 ndash ndash 083Hkfrac1415 025 020 036 037 028 ndash ndash 017Hkfrac1420 047 044 071 077 054 ndash ndash 036

(nfrac14 40 alphafrac14 010)Fkfrac1415 060 070 043 043 ndash 089 040 073Fkfrac1420 098 099 091 089 ndash 100 089 099Gkfrac1415 019 022 057 061 ndash ndash 039 023Gkfrac1420 056 072 096 098 ndash ndash 085 075Gkfrac1430 ndash ndash ndash ndash ndash ndash ndash 100Hkfrac1415 036 032 058 063 ndash ndash ndash 027Hkfrac1420 071 080 096 098 ndash ndash ndash 075

192 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Hollander M Wolfe D A (1973) Nonparametric Statistical Methods2nd ed New York Wiley

Stephens M A (1974) EDF statistics for goodness-of-fit and somecomparisons (In theory and methods) J Amer Stat Assoc Theoryand Methods Section September 69(347)730ndash737

A New One-Sample Test for Goodness-of-Fit 193

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Request PermissionOrder Reprints

Reprints of this article can also be ordered at

httpwwwdekkercomservletproductDOI101081STA120026585

Request Permission or Order Reprints Instantly

Interested in copying and sharing this article In most cases US Copyright Law requires that you get permission from the articlersquos rightsholder before using copyrighted content

All information and materials found in this article including but not limited to text trademarks patents logos graphics and images (the Materials) are the copyrighted works and other forms of intellectual property of Marcel Dekker Inc or its licensors All rights not expressly granted are reserved

Get permission to lawfully reproduce and distribute the Materials or order reprints quickly and painlessly Simply click on the Request Permission Order Reprints link below and follow the instructions Visit the US Copyright Office for information on Fair Use limitations of US copyright law Please refer to The Association of American Publishersrsquo (AAP) website for guidelines on Fair Use in the Classroom

The Materials are for your personal use only and cannot be reformatted reposted resold or distributed by electronic means or otherwise without permission from Marcel Dekker Inc Marcel Dekker Inc grants you the limited right to display the Materials only on your personal computer or personal wireless device and to copy and download single copies of such Materials provided that any copyright trademark or other notice appearing on such Materials is also retained by displayed copied or downloaded as part of the Materials and is not removed or obscured and provided you do not edit modify alter or enhance the Materials Please refer to our Website User Agreement for more details

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Page 8: A New One-Sample Test for Goodness-of-Fit

ORDER REPRINTS

In each case the four alpha levels were selected to be near 020 010 005and 001

Here are some additional critical values for values of n from 8through 15 (Tables 5 and 6)

4 TABLES COMPARING POWER OF THE

A STATISTIC WITH OTHER

ONE-SAMPLE TESTS

As noted earlier the table of results comparing Kolmogorov-Smirnov D Cramer-von Mises W2 Kuiper V Watson U2 Anderson-Darling A2 Q (frac14Si ln zi) and Chi2 first appeared in an article byStephens (1974) in JASA I have simply added the column showing thepower of the A statistic as estimated by a number of simulations

Table 3

A nfrac14 2 nfrac14 3 nfrac14 4 nfrac14 5 nfrac14 6 nfrac14 7

0 0500 022222 009375 003840 001543 000611991 1000 066667 037500 019200 009259 004283932 092593 068750 044160 025720 013973773 100000 087500 067200 046296 029273524 096094 082560 064472 046256235 099219 091520 077761 061317996 100000 096352 086703 073131097 098656 092438 081845158 099616 095945 088053459 099936 097990 0923662710 100000 099096 0952885311 099640 0972077912 099880 0984233913 099970 0991597214 099996 0995813215 100000 0998074216 0999198617 0999708518 0999912619 0999980620 0999997621 10000000

A New One-Sample Test for Goodness-of-Fit 187

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

10000 8 9 02267 1000 18 32 019611 01183 38 010113 00543 44 004917 00089 57 0009

10000 9 11 02074 1000 19 33 020413 01176 40 010216 00462 48 004920 00106 63 0010

10000 10 13 02008 10000 20 37 0197316 00972 45 0099319 00406 53 0047724 00090 66 00092

10000 11 15 02018 1000 21 40 019719 00860 49 009521 00522 56 005127 00111 75 0009

10000 12 17 02015 1000 22 43 020020 01162 51 010024 00507 59 004930 00099 77 0010

10000 13 19 02056 1000 23 47 019823 01038 56 010027 00484 66 004935 00097 83 0010

10000 14 22 01909 1200 24 48 0200827 00945 58 0098331 00497 67 0049240 00085 87 00092

10000 15 23 02125 10000 25 51 0199129 00972 61 0103134 00468 72 0051543 00089 93 00107

1000 16 26 0192 1600 26 53 0203131 0092 65 0101336 0046 77 0048148 0010 97 00094

1000 17 29 0193 1600 27 57 0196934 0102 69 0098840 0045 80 0047551 0008 104 00100

188 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4 Continued

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

1600 28 59 01981 1600 38 96 02019

73 01019 118 0101386 00500 135 00500110 00100 181 00100

1600 29 66 01975 1600 39 99 0201380 01013 120 0100092 00481 140 00494117 00100 170 00094

10000 30 67 02029 10000 40 103 0201181 01025 126 0100095 00490 148 00496126 00095 191 00099

1600 31 71 01969 10000 50 145 0194587 00981 177 00984101 00500 205 00496128 00100 265 00099

1600 32 74 02038 10000 60 190 0199991 01000 232 00986107 00488 270 00498135 00100 348 00093

1600 33 77 01963 10000 70 240 0201990 01006 292 01027108 00500 340 00488141 00094 438 00095

1600 34 82 01975 10000 80 293 01988101 00988 357 01015117 00488 415 00530149 00100 535 00112

10000 35 84 02036 10000 90 349 01982102 01019 426 01015120 00493 495 00530152 00103 639 00112

1600 36 91 02019 10000 100 409 01930113 00981 499 00955132 00488 580 00467166 00100 748 00089

1600 37 90 01988110 01000128 00488172 00100

A New One-Sample Test for Goodness-of-Fit 189

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

5

nfrac148

nfrac149

nfrac1410

nfrac1411

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

902267

11

02074

13

02008

15

02018

10

01657

12

01580

14

01606

16

01628

11

01183

13

01176

15

01260

17

01336

12

00810

14

00874

16

00972

18

01091

13

00543

15

00633

17

00729

19

00860

14

00347

16

00462

18

00542

20

00670

15

00227

17

00335

19

00406

21

00522

16

00151

18

00241

20

00298

22

00417

17

00089

19

00157

21

00221

23

00334

20

00106

22

00166

24

00257

21

00062

23

00126

25

00194

24

00090

26

00147

27

00111

28

00082

190 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

6

nfrac1412

nfrac1413

nfrac1414

nfrac1415

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

17

02015

19

02056

21

02168

23

02125

18

01677

20

01733

22

01909

24

01873

19

01407

21

01488

23

01657

25

01648

20

01162

22

01249

24

01431

26

01454

21

00929

23

01038

25

01267

27

01285

22

00755

24

00878

26

01100

28

01132

23

00615

25

00736

27

00945

29

00972

24

00507

26

00606

28

00806

30

00841

25

00416

27

00484

29

00691

31

00723

26

00315

28

00410

30

00581

32

00624

27

00240

29

00334

31

00497

33

00542

28

00175

30

00278

32

00435

34

00468

29

00134

31

00221

33

00360

35

00403

30

00099

32

00184

34

00293

36

00344

33

00145

35

00236

37

00296

34

00120

36

00200

38

00240

35

00097

37

00175

39

00206

38

00141

40

00163

39

00119

41

00141

40

00085

42

00118

43

00089

A New One-Sample Test for Goodness-of-Fit 191

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

(Table 7) The alternative distribution functions are described in sectiontwo of this paper The sample sizes (n) were set equal to 10 20 and 40The alpha level was set at 010

REFERENCES

Bradley J V (1968) Distribution-Free Statistical Tests EnglewoodCliffs NJ Prentice-Hall

Gibbons J D (1976) Nonparametric Methods for Quantitative AnalysisNew York Holt Rinehart and Winston

Gibbons J D (1985) Nonparametric Statistical Inference New YorkMarcel Dekker

Table 7

Alternative D W2 V U2 A2 Q Chi2 A

(nfrac14 10 alphafrac14 010)Fkfrac1415 023 027 018 019 024 043 ndash 027Fkfrac1420 054 060 035 035 058 ndash ndash 058Gkfrac1415 009 007 022 023 006 ndash ndash 007Gkfrac1420 009 007 040 044 006 ndash ndash 010Gkfrac1430 021 021 081 086 018 ndash ndash 029Hkfrac1415 ndash ndash ndash ndash ndash ndash ndash 014Hkfrac1420 ndash ndash ndash ndash ndash ndash ndash 021

(nfrac14 20 alphafrac14 010)Fkfrac1415 038 046 025 028 046 068 ndash 046Fkfrac1420 078 087 061 060 087 097 059 088Gkfrac1415 013 011 032 034 010 011 ndash 010Gkfrac1420 025 025 071 077 028 025 ndash 029Gkfrac1430 063 079 099 099 084 ndash ndash 083Hkfrac1415 025 020 036 037 028 ndash ndash 017Hkfrac1420 047 044 071 077 054 ndash ndash 036

(nfrac14 40 alphafrac14 010)Fkfrac1415 060 070 043 043 ndash 089 040 073Fkfrac1420 098 099 091 089 ndash 100 089 099Gkfrac1415 019 022 057 061 ndash ndash 039 023Gkfrac1420 056 072 096 098 ndash ndash 085 075Gkfrac1430 ndash ndash ndash ndash ndash ndash ndash 100Hkfrac1415 036 032 058 063 ndash ndash ndash 027Hkfrac1420 071 080 096 098 ndash ndash ndash 075

192 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Hollander M Wolfe D A (1973) Nonparametric Statistical Methods2nd ed New York Wiley

Stephens M A (1974) EDF statistics for goodness-of-fit and somecomparisons (In theory and methods) J Amer Stat Assoc Theoryand Methods Section September 69(347)730ndash737

A New One-Sample Test for Goodness-of-Fit 193

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Request PermissionOrder Reprints

Reprints of this article can also be ordered at

httpwwwdekkercomservletproductDOI101081STA120026585

Request Permission or Order Reprints Instantly

Interested in copying and sharing this article In most cases US Copyright Law requires that you get permission from the articlersquos rightsholder before using copyrighted content

All information and materials found in this article including but not limited to text trademarks patents logos graphics and images (the Materials) are the copyrighted works and other forms of intellectual property of Marcel Dekker Inc or its licensors All rights not expressly granted are reserved

Get permission to lawfully reproduce and distribute the Materials or order reprints quickly and painlessly Simply click on the Request Permission Order Reprints link below and follow the instructions Visit the US Copyright Office for information on Fair Use limitations of US copyright law Please refer to The Association of American Publishersrsquo (AAP) website for guidelines on Fair Use in the Classroom

The Materials are for your personal use only and cannot be reformatted reposted resold or distributed by electronic means or otherwise without permission from Marcel Dekker Inc Marcel Dekker Inc grants you the limited right to display the Materials only on your personal computer or personal wireless device and to copy and download single copies of such Materials provided that any copyright trademark or other notice appearing on such Materials is also retained by displayed copied or downloaded as part of the Materials and is not removed or obscured and provided you do not edit modify alter or enhance the Materials Please refer to our Website User Agreement for more details

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Page 9: A New One-Sample Test for Goodness-of-Fit

ORDER REPRINTS

Table 4

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

10000 8 9 02267 1000 18 32 019611 01183 38 010113 00543 44 004917 00089 57 0009

10000 9 11 02074 1000 19 33 020413 01176 40 010216 00462 48 004920 00106 63 0010

10000 10 13 02008 10000 20 37 0197316 00972 45 0099319 00406 53 0047724 00090 66 00092

10000 11 15 02018 1000 21 40 019719 00860 49 009521 00522 56 005127 00111 75 0009

10000 12 17 02015 1000 22 43 020020 01162 51 010024 00507 59 004930 00099 77 0010

10000 13 19 02056 1000 23 47 019823 01038 56 010027 00484 66 004935 00097 83 0010

10000 14 22 01909 1200 24 48 0200827 00945 58 0098331 00497 67 0049240 00085 87 00092

10000 15 23 02125 10000 25 51 0199129 00972 61 0103134 00468 72 0051543 00089 93 00107

1000 16 26 0192 1600 26 53 0203131 0092 65 0101336 0046 77 0048148 0010 97 00094

1000 17 29 0193 1600 27 57 0196934 0102 69 0098840 0045 80 0047551 0008 104 00100

188 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table 4 Continued

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

1600 28 59 01981 1600 38 96 02019

73 01019 118 0101386 00500 135 00500110 00100 181 00100

1600 29 66 01975 1600 39 99 0201380 01013 120 0100092 00481 140 00494117 00100 170 00094

10000 30 67 02029 10000 40 103 0201181 01025 126 0100095 00490 148 00496126 00095 191 00099

1600 31 71 01969 10000 50 145 0194587 00981 177 00984101 00500 205 00496128 00100 265 00099

1600 32 74 02038 10000 60 190 0199991 01000 232 00986107 00488 270 00498135 00100 348 00093

1600 33 77 01963 10000 70 240 0201990 01006 292 01027108 00500 340 00488141 00094 438 00095

1600 34 82 01975 10000 80 293 01988101 00988 357 01015117 00488 415 00530149 00100 535 00112

10000 35 84 02036 10000 90 349 01982102 01019 426 01015120 00493 495 00530152 00103 639 00112

1600 36 91 02019 10000 100 409 01930113 00981 499 00955132 00488 580 00467166 00100 748 00089

1600 37 90 01988110 01000128 00488172 00100

A New One-Sample Test for Goodness-of-Fit 189

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

5

nfrac148

nfrac149

nfrac1410

nfrac1411

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

902267

11

02074

13

02008

15

02018

10

01657

12

01580

14

01606

16

01628

11

01183

13

01176

15

01260

17

01336

12

00810

14

00874

16

00972

18

01091

13

00543

15

00633

17

00729

19

00860

14

00347

16

00462

18

00542

20

00670

15

00227

17

00335

19

00406

21

00522

16

00151

18

00241

20

00298

22

00417

17

00089

19

00157

21

00221

23

00334

20

00106

22

00166

24

00257

21

00062

23

00126

25

00194

24

00090

26

00147

27

00111

28

00082

190 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

6

nfrac1412

nfrac1413

nfrac1414

nfrac1415

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

17

02015

19

02056

21

02168

23

02125

18

01677

20

01733

22

01909

24

01873

19

01407

21

01488

23

01657

25

01648

20

01162

22

01249

24

01431

26

01454

21

00929

23

01038

25

01267

27

01285

22

00755

24

00878

26

01100

28

01132

23

00615

25

00736

27

00945

29

00972

24

00507

26

00606

28

00806

30

00841

25

00416

27

00484

29

00691

31

00723

26

00315

28

00410

30

00581

32

00624

27

00240

29

00334

31

00497

33

00542

28

00175

30

00278

32

00435

34

00468

29

00134

31

00221

33

00360

35

00403

30

00099

32

00184

34

00293

36

00344

33

00145

35

00236

37

00296

34

00120

36

00200

38

00240

35

00097

37

00175

39

00206

38

00141

40

00163

39

00119

41

00141

40

00085

42

00118

43

00089

A New One-Sample Test for Goodness-of-Fit 191

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

(Table 7) The alternative distribution functions are described in sectiontwo of this paper The sample sizes (n) were set equal to 10 20 and 40The alpha level was set at 010

REFERENCES

Bradley J V (1968) Distribution-Free Statistical Tests EnglewoodCliffs NJ Prentice-Hall

Gibbons J D (1976) Nonparametric Methods for Quantitative AnalysisNew York Holt Rinehart and Winston

Gibbons J D (1985) Nonparametric Statistical Inference New YorkMarcel Dekker

Table 7

Alternative D W2 V U2 A2 Q Chi2 A

(nfrac14 10 alphafrac14 010)Fkfrac1415 023 027 018 019 024 043 ndash 027Fkfrac1420 054 060 035 035 058 ndash ndash 058Gkfrac1415 009 007 022 023 006 ndash ndash 007Gkfrac1420 009 007 040 044 006 ndash ndash 010Gkfrac1430 021 021 081 086 018 ndash ndash 029Hkfrac1415 ndash ndash ndash ndash ndash ndash ndash 014Hkfrac1420 ndash ndash ndash ndash ndash ndash ndash 021

(nfrac14 20 alphafrac14 010)Fkfrac1415 038 046 025 028 046 068 ndash 046Fkfrac1420 078 087 061 060 087 097 059 088Gkfrac1415 013 011 032 034 010 011 ndash 010Gkfrac1420 025 025 071 077 028 025 ndash 029Gkfrac1430 063 079 099 099 084 ndash ndash 083Hkfrac1415 025 020 036 037 028 ndash ndash 017Hkfrac1420 047 044 071 077 054 ndash ndash 036

(nfrac14 40 alphafrac14 010)Fkfrac1415 060 070 043 043 ndash 089 040 073Fkfrac1420 098 099 091 089 ndash 100 089 099Gkfrac1415 019 022 057 061 ndash ndash 039 023Gkfrac1420 056 072 096 098 ndash ndash 085 075Gkfrac1430 ndash ndash ndash ndash ndash ndash ndash 100Hkfrac1415 036 032 058 063 ndash ndash ndash 027Hkfrac1420 071 080 096 098 ndash ndash ndash 075

192 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Hollander M Wolfe D A (1973) Nonparametric Statistical Methods2nd ed New York Wiley

Stephens M A (1974) EDF statistics for goodness-of-fit and somecomparisons (In theory and methods) J Amer Stat Assoc Theoryand Methods Section September 69(347)730ndash737

A New One-Sample Test for Goodness-of-Fit 193

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Request PermissionOrder Reprints

Reprints of this article can also be ordered at

httpwwwdekkercomservletproductDOI101081STA120026585

Request Permission or Order Reprints Instantly

Interested in copying and sharing this article In most cases US Copyright Law requires that you get permission from the articlersquos rightsholder before using copyrighted content

All information and materials found in this article including but not limited to text trademarks patents logos graphics and images (the Materials) are the copyrighted works and other forms of intellectual property of Marcel Dekker Inc or its licensors All rights not expressly granted are reserved

Get permission to lawfully reproduce and distribute the Materials or order reprints quickly and painlessly Simply click on the Request Permission Order Reprints link below and follow the instructions Visit the US Copyright Office for information on Fair Use limitations of US copyright law Please refer to The Association of American Publishersrsquo (AAP) website for guidelines on Fair Use in the Classroom

The Materials are for your personal use only and cannot be reformatted reposted resold or distributed by electronic means or otherwise without permission from Marcel Dekker Inc Marcel Dekker Inc grants you the limited right to display the Materials only on your personal computer or personal wireless device and to copy and download single copies of such Materials provided that any copyright trademark or other notice appearing on such Materials is also retained by displayed copied or downloaded as part of the Materials and is not removed or obscured and provided you do not edit modify alter or enhance the Materials Please refer to our Website User Agreement for more details

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Page 10: A New One-Sample Test for Goodness-of-Fit

ORDER REPRINTS

Table 4 Continued

N nCritical

value (A) P(A) gtfrac14A N nCritical

value (A) P(A) gtfrac14A

1600 28 59 01981 1600 38 96 02019

73 01019 118 0101386 00500 135 00500110 00100 181 00100

1600 29 66 01975 1600 39 99 0201380 01013 120 0100092 00481 140 00494117 00100 170 00094

10000 30 67 02029 10000 40 103 0201181 01025 126 0100095 00490 148 00496126 00095 191 00099

1600 31 71 01969 10000 50 145 0194587 00981 177 00984101 00500 205 00496128 00100 265 00099

1600 32 74 02038 10000 60 190 0199991 01000 232 00986107 00488 270 00498135 00100 348 00093

1600 33 77 01963 10000 70 240 0201990 01006 292 01027108 00500 340 00488141 00094 438 00095

1600 34 82 01975 10000 80 293 01988101 00988 357 01015117 00488 415 00530149 00100 535 00112

10000 35 84 02036 10000 90 349 01982102 01019 426 01015120 00493 495 00530152 00103 639 00112

1600 36 91 02019 10000 100 409 01930113 00981 499 00955132 00488 580 00467166 00100 748 00089

1600 37 90 01988110 01000128 00488172 00100

A New One-Sample Test for Goodness-of-Fit 189

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

5

nfrac148

nfrac149

nfrac1410

nfrac1411

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

902267

11

02074

13

02008

15

02018

10

01657

12

01580

14

01606

16

01628

11

01183

13

01176

15

01260

17

01336

12

00810

14

00874

16

00972

18

01091

13

00543

15

00633

17

00729

19

00860

14

00347

16

00462

18

00542

20

00670

15

00227

17

00335

19

00406

21

00522

16

00151

18

00241

20

00298

22

00417

17

00089

19

00157

21

00221

23

00334

20

00106

22

00166

24

00257

21

00062

23

00126

25

00194

24

00090

26

00147

27

00111

28

00082

190 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

6

nfrac1412

nfrac1413

nfrac1414

nfrac1415

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

17

02015

19

02056

21

02168

23

02125

18

01677

20

01733

22

01909

24

01873

19

01407

21

01488

23

01657

25

01648

20

01162

22

01249

24

01431

26

01454

21

00929

23

01038

25

01267

27

01285

22

00755

24

00878

26

01100

28

01132

23

00615

25

00736

27

00945

29

00972

24

00507

26

00606

28

00806

30

00841

25

00416

27

00484

29

00691

31

00723

26

00315

28

00410

30

00581

32

00624

27

00240

29

00334

31

00497

33

00542

28

00175

30

00278

32

00435

34

00468

29

00134

31

00221

33

00360

35

00403

30

00099

32

00184

34

00293

36

00344

33

00145

35

00236

37

00296

34

00120

36

00200

38

00240

35

00097

37

00175

39

00206

38

00141

40

00163

39

00119

41

00141

40

00085

42

00118

43

00089

A New One-Sample Test for Goodness-of-Fit 191

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

(Table 7) The alternative distribution functions are described in sectiontwo of this paper The sample sizes (n) were set equal to 10 20 and 40The alpha level was set at 010

REFERENCES

Bradley J V (1968) Distribution-Free Statistical Tests EnglewoodCliffs NJ Prentice-Hall

Gibbons J D (1976) Nonparametric Methods for Quantitative AnalysisNew York Holt Rinehart and Winston

Gibbons J D (1985) Nonparametric Statistical Inference New YorkMarcel Dekker

Table 7

Alternative D W2 V U2 A2 Q Chi2 A

(nfrac14 10 alphafrac14 010)Fkfrac1415 023 027 018 019 024 043 ndash 027Fkfrac1420 054 060 035 035 058 ndash ndash 058Gkfrac1415 009 007 022 023 006 ndash ndash 007Gkfrac1420 009 007 040 044 006 ndash ndash 010Gkfrac1430 021 021 081 086 018 ndash ndash 029Hkfrac1415 ndash ndash ndash ndash ndash ndash ndash 014Hkfrac1420 ndash ndash ndash ndash ndash ndash ndash 021

(nfrac14 20 alphafrac14 010)Fkfrac1415 038 046 025 028 046 068 ndash 046Fkfrac1420 078 087 061 060 087 097 059 088Gkfrac1415 013 011 032 034 010 011 ndash 010Gkfrac1420 025 025 071 077 028 025 ndash 029Gkfrac1430 063 079 099 099 084 ndash ndash 083Hkfrac1415 025 020 036 037 028 ndash ndash 017Hkfrac1420 047 044 071 077 054 ndash ndash 036

(nfrac14 40 alphafrac14 010)Fkfrac1415 060 070 043 043 ndash 089 040 073Fkfrac1420 098 099 091 089 ndash 100 089 099Gkfrac1415 019 022 057 061 ndash ndash 039 023Gkfrac1420 056 072 096 098 ndash ndash 085 075Gkfrac1430 ndash ndash ndash ndash ndash ndash ndash 100Hkfrac1415 036 032 058 063 ndash ndash ndash 027Hkfrac1420 071 080 096 098 ndash ndash ndash 075

192 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Hollander M Wolfe D A (1973) Nonparametric Statistical Methods2nd ed New York Wiley

Stephens M A (1974) EDF statistics for goodness-of-fit and somecomparisons (In theory and methods) J Amer Stat Assoc Theoryand Methods Section September 69(347)730ndash737

A New One-Sample Test for Goodness-of-Fit 193

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Request PermissionOrder Reprints

Reprints of this article can also be ordered at

httpwwwdekkercomservletproductDOI101081STA120026585

Request Permission or Order Reprints Instantly

Interested in copying and sharing this article In most cases US Copyright Law requires that you get permission from the articlersquos rightsholder before using copyrighted content

All information and materials found in this article including but not limited to text trademarks patents logos graphics and images (the Materials) are the copyrighted works and other forms of intellectual property of Marcel Dekker Inc or its licensors All rights not expressly granted are reserved

Get permission to lawfully reproduce and distribute the Materials or order reprints quickly and painlessly Simply click on the Request Permission Order Reprints link below and follow the instructions Visit the US Copyright Office for information on Fair Use limitations of US copyright law Please refer to The Association of American Publishersrsquo (AAP) website for guidelines on Fair Use in the Classroom

The Materials are for your personal use only and cannot be reformatted reposted resold or distributed by electronic means or otherwise without permission from Marcel Dekker Inc Marcel Dekker Inc grants you the limited right to display the Materials only on your personal computer or personal wireless device and to copy and download single copies of such Materials provided that any copyright trademark or other notice appearing on such Materials is also retained by displayed copied or downloaded as part of the Materials and is not removed or obscured and provided you do not edit modify alter or enhance the Materials Please refer to our Website User Agreement for more details

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Page 11: A New One-Sample Test for Goodness-of-Fit

ORDER REPRINTS

Table

5

nfrac148

nfrac149

nfrac1410

nfrac1411

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

902267

11

02074

13

02008

15

02018

10

01657

12

01580

14

01606

16

01628

11

01183

13

01176

15

01260

17

01336

12

00810

14

00874

16

00972

18

01091

13

00543

15

00633

17

00729

19

00860

14

00347

16

00462

18

00542

20

00670

15

00227

17

00335

19

00406

21

00522

16

00151

18

00241

20

00298

22

00417

17

00089

19

00157

21

00221

23

00334

20

00106

22

00166

24

00257

21

00062

23

00126

25

00194

24

00090

26

00147

27

00111

28

00082

190 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Table

6

nfrac1412

nfrac1413

nfrac1414

nfrac1415

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

17

02015

19

02056

21

02168

23

02125

18

01677

20

01733

22

01909

24

01873

19

01407

21

01488

23

01657

25

01648

20

01162

22

01249

24

01431

26

01454

21

00929

23

01038

25

01267

27

01285

22

00755

24

00878

26

01100

28

01132

23

00615

25

00736

27

00945

29

00972

24

00507

26

00606

28

00806

30

00841

25

00416

27

00484

29

00691

31

00723

26

00315

28

00410

30

00581

32

00624

27

00240

29

00334

31

00497

33

00542

28

00175

30

00278

32

00435

34

00468

29

00134

31

00221

33

00360

35

00403

30

00099

32

00184

34

00293

36

00344

33

00145

35

00236

37

00296

34

00120

36

00200

38

00240

35

00097

37

00175

39

00206

38

00141

40

00163

39

00119

41

00141

40

00085

42

00118

43

00089

A New One-Sample Test for Goodness-of-Fit 191

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

(Table 7) The alternative distribution functions are described in sectiontwo of this paper The sample sizes (n) were set equal to 10 20 and 40The alpha level was set at 010

REFERENCES

Bradley J V (1968) Distribution-Free Statistical Tests EnglewoodCliffs NJ Prentice-Hall

Gibbons J D (1976) Nonparametric Methods for Quantitative AnalysisNew York Holt Rinehart and Winston

Gibbons J D (1985) Nonparametric Statistical Inference New YorkMarcel Dekker

Table 7

Alternative D W2 V U2 A2 Q Chi2 A

(nfrac14 10 alphafrac14 010)Fkfrac1415 023 027 018 019 024 043 ndash 027Fkfrac1420 054 060 035 035 058 ndash ndash 058Gkfrac1415 009 007 022 023 006 ndash ndash 007Gkfrac1420 009 007 040 044 006 ndash ndash 010Gkfrac1430 021 021 081 086 018 ndash ndash 029Hkfrac1415 ndash ndash ndash ndash ndash ndash ndash 014Hkfrac1420 ndash ndash ndash ndash ndash ndash ndash 021

(nfrac14 20 alphafrac14 010)Fkfrac1415 038 046 025 028 046 068 ndash 046Fkfrac1420 078 087 061 060 087 097 059 088Gkfrac1415 013 011 032 034 010 011 ndash 010Gkfrac1420 025 025 071 077 028 025 ndash 029Gkfrac1430 063 079 099 099 084 ndash ndash 083Hkfrac1415 025 020 036 037 028 ndash ndash 017Hkfrac1420 047 044 071 077 054 ndash ndash 036

(nfrac14 40 alphafrac14 010)Fkfrac1415 060 070 043 043 ndash 089 040 073Fkfrac1420 098 099 091 089 ndash 100 089 099Gkfrac1415 019 022 057 061 ndash ndash 039 023Gkfrac1420 056 072 096 098 ndash ndash 085 075Gkfrac1430 ndash ndash ndash ndash ndash ndash ndash 100Hkfrac1415 036 032 058 063 ndash ndash ndash 027Hkfrac1420 071 080 096 098 ndash ndash ndash 075

192 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Hollander M Wolfe D A (1973) Nonparametric Statistical Methods2nd ed New York Wiley

Stephens M A (1974) EDF statistics for goodness-of-fit and somecomparisons (In theory and methods) J Amer Stat Assoc Theoryand Methods Section September 69(347)730ndash737

A New One-Sample Test for Goodness-of-Fit 193

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Request PermissionOrder Reprints

Reprints of this article can also be ordered at

httpwwwdekkercomservletproductDOI101081STA120026585

Request Permission or Order Reprints Instantly

Interested in copying and sharing this article In most cases US Copyright Law requires that you get permission from the articlersquos rightsholder before using copyrighted content

All information and materials found in this article including but not limited to text trademarks patents logos graphics and images (the Materials) are the copyrighted works and other forms of intellectual property of Marcel Dekker Inc or its licensors All rights not expressly granted are reserved

Get permission to lawfully reproduce and distribute the Materials or order reprints quickly and painlessly Simply click on the Request Permission Order Reprints link below and follow the instructions Visit the US Copyright Office for information on Fair Use limitations of US copyright law Please refer to The Association of American Publishersrsquo (AAP) website for guidelines on Fair Use in the Classroom

The Materials are for your personal use only and cannot be reformatted reposted resold or distributed by electronic means or otherwise without permission from Marcel Dekker Inc Marcel Dekker Inc grants you the limited right to display the Materials only on your personal computer or personal wireless device and to copy and download single copies of such Materials provided that any copyright trademark or other notice appearing on such Materials is also retained by displayed copied or downloaded as part of the Materials and is not removed or obscured and provided you do not edit modify alter or enhance the Materials Please refer to our Website User Agreement for more details

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Page 12: A New One-Sample Test for Goodness-of-Fit

ORDER REPRINTS

Table

6

nfrac1412

nfrac1413

nfrac1414

nfrac1415

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

Critical

value(A

)P(A

)gtfrac14A

17

02015

19

02056

21

02168

23

02125

18

01677

20

01733

22

01909

24

01873

19

01407

21

01488

23

01657

25

01648

20

01162

22

01249

24

01431

26

01454

21

00929

23

01038

25

01267

27

01285

22

00755

24

00878

26

01100

28

01132

23

00615

25

00736

27

00945

29

00972

24

00507

26

00606

28

00806

30

00841

25

00416

27

00484

29

00691

31

00723

26

00315

28

00410

30

00581

32

00624

27

00240

29

00334

31

00497

33

00542

28

00175

30

00278

32

00435

34

00468

29

00134

31

00221

33

00360

35

00403

30

00099

32

00184

34

00293

36

00344

33

00145

35

00236

37

00296

34

00120

36

00200

38

00240

35

00097

37

00175

39

00206

38

00141

40

00163

39

00119

41

00141

40

00085

42

00118

43

00089

A New One-Sample Test for Goodness-of-Fit 191

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

(Table 7) The alternative distribution functions are described in sectiontwo of this paper The sample sizes (n) were set equal to 10 20 and 40The alpha level was set at 010

REFERENCES

Bradley J V (1968) Distribution-Free Statistical Tests EnglewoodCliffs NJ Prentice-Hall

Gibbons J D (1976) Nonparametric Methods for Quantitative AnalysisNew York Holt Rinehart and Winston

Gibbons J D (1985) Nonparametric Statistical Inference New YorkMarcel Dekker

Table 7

Alternative D W2 V U2 A2 Q Chi2 A

(nfrac14 10 alphafrac14 010)Fkfrac1415 023 027 018 019 024 043 ndash 027Fkfrac1420 054 060 035 035 058 ndash ndash 058Gkfrac1415 009 007 022 023 006 ndash ndash 007Gkfrac1420 009 007 040 044 006 ndash ndash 010Gkfrac1430 021 021 081 086 018 ndash ndash 029Hkfrac1415 ndash ndash ndash ndash ndash ndash ndash 014Hkfrac1420 ndash ndash ndash ndash ndash ndash ndash 021

(nfrac14 20 alphafrac14 010)Fkfrac1415 038 046 025 028 046 068 ndash 046Fkfrac1420 078 087 061 060 087 097 059 088Gkfrac1415 013 011 032 034 010 011 ndash 010Gkfrac1420 025 025 071 077 028 025 ndash 029Gkfrac1430 063 079 099 099 084 ndash ndash 083Hkfrac1415 025 020 036 037 028 ndash ndash 017Hkfrac1420 047 044 071 077 054 ndash ndash 036

(nfrac14 40 alphafrac14 010)Fkfrac1415 060 070 043 043 ndash 089 040 073Fkfrac1420 098 099 091 089 ndash 100 089 099Gkfrac1415 019 022 057 061 ndash ndash 039 023Gkfrac1420 056 072 096 098 ndash ndash 085 075Gkfrac1430 ndash ndash ndash ndash ndash ndash ndash 100Hkfrac1415 036 032 058 063 ndash ndash ndash 027Hkfrac1420 071 080 096 098 ndash ndash ndash 075

192 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Hollander M Wolfe D A (1973) Nonparametric Statistical Methods2nd ed New York Wiley

Stephens M A (1974) EDF statistics for goodness-of-fit and somecomparisons (In theory and methods) J Amer Stat Assoc Theoryand Methods Section September 69(347)730ndash737

A New One-Sample Test for Goodness-of-Fit 193

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Request PermissionOrder Reprints

Reprints of this article can also be ordered at

httpwwwdekkercomservletproductDOI101081STA120026585

Request Permission or Order Reprints Instantly

Interested in copying and sharing this article In most cases US Copyright Law requires that you get permission from the articlersquos rightsholder before using copyrighted content

All information and materials found in this article including but not limited to text trademarks patents logos graphics and images (the Materials) are the copyrighted works and other forms of intellectual property of Marcel Dekker Inc or its licensors All rights not expressly granted are reserved

Get permission to lawfully reproduce and distribute the Materials or order reprints quickly and painlessly Simply click on the Request Permission Order Reprints link below and follow the instructions Visit the US Copyright Office for information on Fair Use limitations of US copyright law Please refer to The Association of American Publishersrsquo (AAP) website for guidelines on Fair Use in the Classroom

The Materials are for your personal use only and cannot be reformatted reposted resold or distributed by electronic means or otherwise without permission from Marcel Dekker Inc Marcel Dekker Inc grants you the limited right to display the Materials only on your personal computer or personal wireless device and to copy and download single copies of such Materials provided that any copyright trademark or other notice appearing on such Materials is also retained by displayed copied or downloaded as part of the Materials and is not removed or obscured and provided you do not edit modify alter or enhance the Materials Please refer to our Website User Agreement for more details

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Page 13: A New One-Sample Test for Goodness-of-Fit

ORDER REPRINTS

(Table 7) The alternative distribution functions are described in sectiontwo of this paper The sample sizes (n) were set equal to 10 20 and 40The alpha level was set at 010

REFERENCES

Bradley J V (1968) Distribution-Free Statistical Tests EnglewoodCliffs NJ Prentice-Hall

Gibbons J D (1976) Nonparametric Methods for Quantitative AnalysisNew York Holt Rinehart and Winston

Gibbons J D (1985) Nonparametric Statistical Inference New YorkMarcel Dekker

Table 7

Alternative D W2 V U2 A2 Q Chi2 A

(nfrac14 10 alphafrac14 010)Fkfrac1415 023 027 018 019 024 043 ndash 027Fkfrac1420 054 060 035 035 058 ndash ndash 058Gkfrac1415 009 007 022 023 006 ndash ndash 007Gkfrac1420 009 007 040 044 006 ndash ndash 010Gkfrac1430 021 021 081 086 018 ndash ndash 029Hkfrac1415 ndash ndash ndash ndash ndash ndash ndash 014Hkfrac1420 ndash ndash ndash ndash ndash ndash ndash 021

(nfrac14 20 alphafrac14 010)Fkfrac1415 038 046 025 028 046 068 ndash 046Fkfrac1420 078 087 061 060 087 097 059 088Gkfrac1415 013 011 032 034 010 011 ndash 010Gkfrac1420 025 025 071 077 028 025 ndash 029Gkfrac1430 063 079 099 099 084 ndash ndash 083Hkfrac1415 025 020 036 037 028 ndash ndash 017Hkfrac1420 047 044 071 077 054 ndash ndash 036

(nfrac14 40 alphafrac14 010)Fkfrac1415 060 070 043 043 ndash 089 040 073Fkfrac1420 098 099 091 089 ndash 100 089 099Gkfrac1415 019 022 057 061 ndash ndash 039 023Gkfrac1420 056 072 096 098 ndash ndash 085 075Gkfrac1430 ndash ndash ndash ndash ndash ndash ndash 100Hkfrac1415 036 032 058 063 ndash ndash ndash 027Hkfrac1420 071 080 096 098 ndash ndash ndash 075

192 Damico

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

ORDER REPRINTS

Hollander M Wolfe D A (1973) Nonparametric Statistical Methods2nd ed New York Wiley

Stephens M A (1974) EDF statistics for goodness-of-fit and somecomparisons (In theory and methods) J Amer Stat Assoc Theoryand Methods Section September 69(347)730ndash737

A New One-Sample Test for Goodness-of-Fit 193

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Request PermissionOrder Reprints

Reprints of this article can also be ordered at

httpwwwdekkercomservletproductDOI101081STA120026585

Request Permission or Order Reprints Instantly

Interested in copying and sharing this article In most cases US Copyright Law requires that you get permission from the articlersquos rightsholder before using copyrighted content

All information and materials found in this article including but not limited to text trademarks patents logos graphics and images (the Materials) are the copyrighted works and other forms of intellectual property of Marcel Dekker Inc or its licensors All rights not expressly granted are reserved

Get permission to lawfully reproduce and distribute the Materials or order reprints quickly and painlessly Simply click on the Request Permission Order Reprints link below and follow the instructions Visit the US Copyright Office for information on Fair Use limitations of US copyright law Please refer to The Association of American Publishersrsquo (AAP) website for guidelines on Fair Use in the Classroom

The Materials are for your personal use only and cannot be reformatted reposted resold or distributed by electronic means or otherwise without permission from Marcel Dekker Inc Marcel Dekker Inc grants you the limited right to display the Materials only on your personal computer or personal wireless device and to copy and download single copies of such Materials provided that any copyright trademark or other notice appearing on such Materials is also retained by displayed copied or downloaded as part of the Materials and is not removed or obscured and provided you do not edit modify alter or enhance the Materials Please refer to our Website User Agreement for more details

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Page 14: A New One-Sample Test for Goodness-of-Fit

ORDER REPRINTS

Hollander M Wolfe D A (1973) Nonparametric Statistical Methods2nd ed New York Wiley

Stephens M A (1974) EDF statistics for goodness-of-fit and somecomparisons (In theory and methods) J Amer Stat Assoc Theoryand Methods Section September 69(347)730ndash737

A New One-Sample Test for Goodness-of-Fit 193

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Request PermissionOrder Reprints

Reprints of this article can also be ordered at

httpwwwdekkercomservletproductDOI101081STA120026585

Request Permission or Order Reprints Instantly

Interested in copying and sharing this article In most cases US Copyright Law requires that you get permission from the articlersquos rightsholder before using copyrighted content

All information and materials found in this article including but not limited to text trademarks patents logos graphics and images (the Materials) are the copyrighted works and other forms of intellectual property of Marcel Dekker Inc or its licensors All rights not expressly granted are reserved

Get permission to lawfully reproduce and distribute the Materials or order reprints quickly and painlessly Simply click on the Request Permission Order Reprints link below and follow the instructions Visit the US Copyright Office for information on Fair Use limitations of US copyright law Please refer to The Association of American Publishersrsquo (AAP) website for guidelines on Fair Use in the Classroom

The Materials are for your personal use only and cannot be reformatted reposted resold or distributed by electronic means or otherwise without permission from Marcel Dekker Inc Marcel Dekker Inc grants you the limited right to display the Materials only on your personal computer or personal wireless device and to copy and download single copies of such Materials provided that any copyright trademark or other notice appearing on such Materials is also retained by displayed copied or downloaded as part of the Materials and is not removed or obscured and provided you do not edit modify alter or enhance the Materials Please refer to our Website User Agreement for more details

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14

Page 15: A New One-Sample Test for Goodness-of-Fit

Request PermissionOrder Reprints

Reprints of this article can also be ordered at

httpwwwdekkercomservletproductDOI101081STA120026585

Request Permission or Order Reprints Instantly

Interested in copying and sharing this article In most cases US Copyright Law requires that you get permission from the articlersquos rightsholder before using copyrighted content

All information and materials found in this article including but not limited to text trademarks patents logos graphics and images (the Materials) are the copyrighted works and other forms of intellectual property of Marcel Dekker Inc or its licensors All rights not expressly granted are reserved

Get permission to lawfully reproduce and distribute the Materials or order reprints quickly and painlessly Simply click on the Request Permission Order Reprints link below and follow the instructions Visit the US Copyright Office for information on Fair Use limitations of US copyright law Please refer to The Association of American Publishersrsquo (AAP) website for guidelines on Fair Use in the Classroom

The Materials are for your personal use only and cannot be reformatted reposted resold or distributed by electronic means or otherwise without permission from Marcel Dekker Inc Marcel Dekker Inc grants you the limited right to display the Materials only on your personal computer or personal wireless device and to copy and download single copies of such Materials provided that any copyright trademark or other notice appearing on such Materials is also retained by displayed copied or downloaded as part of the Materials and is not removed or obscured and provided you do not edit modify alter or enhance the Materials Please refer to our Website User Agreement for more details

Dow

nloa

ded

by [

Uni

vers

ity o

f C

alif

orni

a R

iver

side

Lib

rari

es]

at 1

005

03

Nov

embe

r 20

14